{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "***\n",
    "***\n",
    "# Topic Modeling Using Turicreate\n",
    "***\n",
    "***\n",
    "\n",
    "王成军\n",
    "\n",
    "wangchengjun@nju.edu.cn\n",
    "\n",
    "计算传播网 http://computational-communication.com"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:10:06.537165Z",
     "start_time": "2019-06-14T15:10:06.534426Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "import turicreate as tc"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Download Data: <del>http://select.cs.cmu.edu/code/graphlab/datasets/wikipedia/wikipedia_raw/w15"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:07.812317Z",
     "start_time": "2019-06-14T15:11:05.552049Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/datalab/bigdata/cjc/w15</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/datalab/bigdata/cjc/w15"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 100 lines in 0.447868 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 100 lines in 0.447868 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "------------------------------------------------------\n",
      "Inferred types from first 100 line(s) of file as \n",
      "column_type_hints=[str]\n",
      "If parsing fails due to incorrect types, you can correct\n",
      "the inferred type list above and pass it to read_csv in\n",
      "the column_type_hints argument\n",
      "------------------------------------------------------\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<pre>Read 12278 lines. Lines per second: 16914.5</pre>"
      ],
      "text/plain": [
       "Read 12278 lines. Lines per second: 16914.5"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Finished parsing file /Users/datalab/bigdata/cjc/w15</pre>"
      ],
      "text/plain": [
       "Finished parsing file /Users/datalab/bigdata/cjc/w15"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Parsing completed. Parsed 72269 lines in 1.73223 secs.</pre>"
      ],
      "text/plain": [
       "Parsing completed. Parsed 72269 lines in 1.73223 secs."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sf = tc.SFrame.read_csv(\"/Users/datalab/bigdata/cjc/w15\", \n",
    "                              header=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:10.863623Z",
     "start_time": "2019-06-14T15:11:10.670898Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\"><table frame=\"box\" rules=\"cols\">\n",
       "    <tr>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">X1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">aynrand born and educated<br>in russia rand migrated ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">asphalt in american<br>english asphalt or ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">actinopterygii the<br>actinopterygii consti ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">altaiclanguages these<br>language families share ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">argon the name argon is<br>derived from the greek ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">augustderleth a 1938<br>guggenheim fellow der ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">amateur amateurism can be<br>seen in both a negative ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">assemblyline an assembly<br>line is a manufacturing ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">astronomicalunit an<br>astronomical unit ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">abbess an abbess latin<br>abbatissa feminine form ...</td>\n",
       "    </tr>\n",
       "</table>\n",
       "[72269 rows x 1 columns]<br/>Note: Only the head of the SFrame is printed.<br/>You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns.\n",
       "</div>"
      ],
      "text/plain": [
       "Columns:\n",
       "\tX1\tstr\n",
       "\n",
       "Rows: 72269\n",
       "\n",
       "Data:\n",
       "+-------------------------------+\n",
       "|               X1              |\n",
       "+-------------------------------+\n",
       "| aynrand born and educated ... |\n",
       "| asphalt in american englis... |\n",
       "| actinopterygii the actinop... |\n",
       "| altaiclanguages these lang... |\n",
       "| argon the name argon is de... |\n",
       "| augustderleth a 1938 gugge... |\n",
       "| amateur amateurism can be ... |\n",
       "| assemblyline an assembly l... |\n",
       "| astronomicalunit an astron... |\n",
       "| abbess an abbess latin abb... |\n",
       "+-------------------------------+\n",
       "[72269 rows x 1 columns]\n",
       "Note: Only the head of the SFrame is printed.\n",
       "You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns."
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Transformations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "https://dato.com/learn/userguide/text/analysis.html"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:14.476922Z",
     "start_time": "2019-06-14T15:11:14.470278Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['_SArray__check_min_observations',\n",
       " '__abs__',\n",
       " '__add__',\n",
       " '__and__',\n",
       " '__bool__',\n",
       " '__class__',\n",
       " '__contains__',\n",
       " '__copy__',\n",
       " '__deepcopy__',\n",
       " '__delattr__',\n",
       " '__dir__',\n",
       " '__div__',\n",
       " '__doc__',\n",
       " '__eq__',\n",
       " '__floordiv__',\n",
       " '__format__',\n",
       " '__ge__',\n",
       " '__get_content_identifier__',\n",
       " '__getattribute__',\n",
       " '__getitem__',\n",
       " '__gt__',\n",
       " '__has_size__',\n",
       " '__hash__',\n",
       " '__init__',\n",
       " '__is_materialized__',\n",
       " '__iter__',\n",
       " '__le__',\n",
       " '__len__',\n",
       " '__lt__',\n",
       " '__materialize__',\n",
       " '__mod__',\n",
       " '__module__',\n",
       " '__mul__',\n",
       " '__ne__',\n",
       " '__neg__',\n",
       " '__new__',\n",
       " '__nonzero__',\n",
       " '__or__',\n",
       " '__pos__',\n",
       " '__pow__',\n",
       " '__proxy__',\n",
       " '__radd__',\n",
       " '__rdiv__',\n",
       " '__reduce__',\n",
       " '__reduce_ex__',\n",
       " '__repr__',\n",
       " '__rfloordiv__',\n",
       " '__rmod__',\n",
       " '__rmul__',\n",
       " '__rpow__',\n",
       " '__rsub__',\n",
       " '__rtruediv__',\n",
       " '__setattr__',\n",
       " '__sizeof__',\n",
       " '__slots__',\n",
       " '__str__',\n",
       " '__sub__',\n",
       " '__subclasshook__',\n",
       " '__truediv__',\n",
       " '_count_ngrams',\n",
       " '_count_words',\n",
       " '_getitem_cache',\n",
       " '_save_as_text',\n",
       " 'all',\n",
       " 'any',\n",
       " 'append',\n",
       " 'apply',\n",
       " 'argmax',\n",
       " 'argmin',\n",
       " 'astype',\n",
       " 'clip',\n",
       " 'clip_lower',\n",
       " 'clip_upper',\n",
       " 'contains',\n",
       " 'countna',\n",
       " 'cumulative_max',\n",
       " 'cumulative_mean',\n",
       " 'cumulative_min',\n",
       " 'cumulative_std',\n",
       " 'cumulative_sum',\n",
       " 'cumulative_var',\n",
       " 'date_range',\n",
       " 'datetime_to_str',\n",
       " 'dict_has_all_keys',\n",
       " 'dict_has_any_keys',\n",
       " 'dict_keys',\n",
       " 'dict_trim_by_keys',\n",
       " 'dict_trim_by_values',\n",
       " 'dict_values',\n",
       " 'dropna',\n",
       " 'dtype',\n",
       " 'element_slice',\n",
       " 'explore',\n",
       " 'fillna',\n",
       " 'filter',\n",
       " 'filter_by',\n",
       " 'from_const',\n",
       " 'from_sequence',\n",
       " 'hash',\n",
       " 'head',\n",
       " 'is_in',\n",
       " 'is_materialized',\n",
       " 'is_topk',\n",
       " 'item_length',\n",
       " 'materialize',\n",
       " 'max',\n",
       " 'mean',\n",
       " 'min',\n",
       " 'nnz',\n",
       " 'pixel_array_to_image',\n",
       " 'plot',\n",
       " 'random_integers',\n",
       " 'random_split',\n",
       " 'read_json',\n",
       " 'rolling_count',\n",
       " 'rolling_max',\n",
       " 'rolling_mean',\n",
       " 'rolling_min',\n",
       " 'rolling_stdv',\n",
       " 'rolling_sum',\n",
       " 'rolling_var',\n",
       " 'sample',\n",
       " 'save',\n",
       " 'shape',\n",
       " 'show',\n",
       " 'sort',\n",
       " 'split_datetime',\n",
       " 'stack',\n",
       " 'std',\n",
       " 'str_to_datetime',\n",
       " 'sum',\n",
       " 'summary',\n",
       " 'tail',\n",
       " 'to_numpy',\n",
       " 'unique',\n",
       " 'unpack',\n",
       " 'value_counts',\n",
       " 'var',\n",
       " 'vector_slice',\n",
       " 'where']"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dir(sf['X1']) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:19.154908Z",
     "start_time": "2019-06-14T15:11:19.150199Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "bow = sf['X1']._count_words() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:19.961430Z",
     "start_time": "2019-06-14T15:11:19.956525Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "turicreate.data_structures.sarray.SArray"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(sf['X1'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:20.710963Z",
     "start_time": "2019-06-14T15:11:20.706859Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "turicreate.data_structures.sarray.SArray"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(bow)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:22.233966Z",
     "start_time": "2019-06-14T15:11:22.116394Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dtype: int\n",
       "Rows: 72269\n",
       "[1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, ... ]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bow.dict_has_any_keys(['limited'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:24.124953Z",
     "start_time": "2019-06-14T15:11:24.109085Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bow.dict_values()[0][:20]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:25.136174Z",
     "start_time": "2019-06-14T15:11:24.947136Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\"><table frame=\"box\" rules=\"cols\">\n",
       "    <tr>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">X1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">aynrand born and educated<br>in russia rand migrated ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">asphalt in american<br>english asphalt or ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">actinopterygii the<br>actinopterygii consti ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">altaiclanguages these<br>language families share ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">argon the name argon is<br>derived from the greek ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">augustderleth a 1938<br>guggenheim fellow der ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">amateur amateurism can be<br>seen in both a negative ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">assemblyline an assembly<br>line is a manufacturing ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">astronomicalunit an<br>astronomical unit ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">abbess an abbess latin<br>abbatissa feminine form ...</td>\n",
       "    </tr>\n",
       "</table>\n",
       "[72269 rows x 1 columns]<br/>Note: Only the head of the SFrame is printed.<br/>You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns.\n",
       "</div>"
      ],
      "text/plain": [
       "Columns:\n",
       "\tX1\tstr\n",
       "\n",
       "Rows: 72269\n",
       "\n",
       "Data:\n",
       "+-------------------------------+\n",
       "|               X1              |\n",
       "+-------------------------------+\n",
       "| aynrand born and educated ... |\n",
       "| asphalt in american englis... |\n",
       "| actinopterygii the actinop... |\n",
       "| altaiclanguages these lang... |\n",
       "| argon the name argon is de... |\n",
       "| augustderleth a 1938 gugge... |\n",
       "| amateur amateurism can be ... |\n",
       "| assemblyline an assembly l... |\n",
       "| astronomicalunit an astron... |\n",
       "| abbess an abbess latin abb... |\n",
       "+-------------------------------+\n",
       "[72269 rows x 1 columns]\n",
       "Note: Only the head of the SFrame is printed.\n",
       "You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns."
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:25.831004Z",
     "start_time": "2019-06-14T15:11:25.827877Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [],
   "source": [
    "sf['bow'] = bow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:27.075901Z",
     "start_time": "2019-06-14T15:11:26.563321Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\"><table frame=\"box\" rules=\"cols\">\n",
       "    <tr>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">X1</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">bow</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">aynrand born and educated<br>in russia rand migrated ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;spoke&#x27;: 1, &#x27;5000&#x27;: 1,<br>&#x27;follows&#x27;: 1, &#x27;given&#x27; ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">asphalt in american<br>english asphalt or ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;lain&#x27;: 1, &#x27;commonly&#x27;:<br>4, &#x27;has&#x27;: 6, &#x27;percent&#x27;: ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">actinopterygii the<br>actinopterygii consti ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;what&#x27;: 1, &#x27;follows&#x27;: 1,<br>&#x27;given&#x27;: 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">altaiclanguages these<br>language families share ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;follows&#x27;: 2, &#x27;has&#x27;: 11,<br>&#x27;general&#x27;: 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">argon the name argon is<br>derived from the greek ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;commonly&#x27;: 1,<br>&#x27;lattice&#x27;: 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">augustderleth a 1938<br>guggenheim fellow der ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;rescue&#x27;: 2, &#x27;amoral&#x27;:<br>1, &#x27;lovecraft&#x27;: 11, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">amateur amateurism can be<br>seen in both a negative ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;receiving&#x27;: 1,<br>&#x27;having&#x27;: 1, &#x27;hand&#x27;: 1, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">assemblyline an assembly<br>line is a manufacturing ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;consider&#x27;: 1, &#x27;world&#x27;:<br>1, &#x27;bring&#x27;: 2, &#x27;pins&#x27; ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">astronomicalunit an<br>astronomical unit ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;given&#x27;: 1, &#x27;ephemeris&#x27;:<br>2, &#x27;world&#x27;: 2, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">abbess an abbess latin<br>abbatissa feminine form ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;major&#x27;: 1, &#x27;abbess&#x27;:<br>10, &#x27;given&#x27;: 1, ...</td>\n",
       "    </tr>\n",
       "</table>\n",
       "[72269 rows x 2 columns]<br/>Note: Only the head of the SFrame is printed.<br/>You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns.\n",
       "</div>"
      ],
      "text/plain": [
       "Columns:\n",
       "\tX1\tstr\n",
       "\tbow\tdict\n",
       "\n",
       "Rows: 72269\n",
       "\n",
       "Data:\n",
       "+-------------------------------+-------------------------------+\n",
       "|               X1              |              bow              |\n",
       "+-------------------------------+-------------------------------+\n",
       "| aynrand born and educated ... | {'spoke': 1, '5000': 1, 'f... |\n",
       "| asphalt in american englis... | {'lain': 1, 'commonly': 4,... |\n",
       "| actinopterygii the actinop... | {'what': 1, 'follows': 1, ... |\n",
       "| altaiclanguages these lang... | {'follows': 2, 'has': 11, ... |\n",
       "| argon the name argon is de... | {'commonly': 1, 'lattice':... |\n",
       "| augustderleth a 1938 gugge... | {'rescue': 2, 'amoral': 1,... |\n",
       "| amateur amateurism can be ... | {'receiving': 1, 'having':... |\n",
       "| assemblyline an assembly l... | {'consider': 1, 'world': 1... |\n",
       "| astronomicalunit an astron... | {'given': 1, 'ephemeris': ... |\n",
       "| abbess an abbess latin abb... | {'major': 1, 'abbess': 10,... |\n",
       "+-------------------------------+-------------------------------+\n",
       "[72269 rows x 2 columns]\n",
       "Note: Only the head of the SFrame is printed.\n",
       "You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns."
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:27.825104Z",
     "start_time": "2019-06-14T15:11:27.820405Z"
    },
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "turicreate.data_structures.sarray.SArray"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(sf['bow'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:28.665108Z",
     "start_time": "2019-06-14T15:11:28.660557Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "72269"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(sf['bow'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:11:43.181325Z",
     "start_time": "2019-06-14T15:11:43.158243Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('spoke', 1), ('5000', 1), ('follows', 1)]"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(sf['bow'][0].items())[:3]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:12:06.137333Z",
     "start_time": "2019-06-14T15:11:48.549415Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "sf['tfidf'] = tc.text_analytics.tf_idf(sf['X1'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:12:09.021826Z",
     "start_time": "2019-06-14T15:12:07.941753Z"
    },
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\"><table frame=\"box\" rules=\"cols\">\n",
       "    <tr>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">X1</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">bow</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">tfidf</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">aynrand born and educated<br>in russia rand migrated ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;spoke&#x27;: 1, &#x27;5000&#x27;: 1,<br>&#x27;follows&#x27;: 1, &#x27;given&#x27; ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;spoke&#x27;:<br>4.830308280673057, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">asphalt in american<br>english asphalt or ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;lain&#x27;: 1, &#x27;commonly&#x27;:<br>4, &#x27;has&#x27;: 6, &#x27;percent&#x27;: ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;lain&#x27;:<br>8.480100346078947, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">actinopterygii the<br>actinopterygii consti ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;what&#x27;: 1, &#x27;follows&#x27;: 1,<br>&#x27;given&#x27;: 1, ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;what&#x27;:<br>2.505781957805935, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">altaiclanguages these<br>language families share ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;follows&#x27;: 2, &#x27;has&#x27;: 11,<br>&#x27;general&#x27;: 1, ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;follows&#x27;:<br>7.5113334785280745, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">argon the name argon is<br>derived from the greek ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;commonly&#x27;: 1,<br>&#x27;lattice&#x27;: 1, ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;commonly&#x27;:<br>3.7717720679882287, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">augustderleth a 1938<br>guggenheim fellow der ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;rescue&#x27;: 2, &#x27;amoral&#x27;:<br>1, &#x27;lovecraft&#x27;: 11, ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;rescue&#x27;:<br>9.19021202607744, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">amateur amateurism can be<br>seen in both a negative ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;receiving&#x27;: 1,<br>&#x27;having&#x27;: 1, &#x27;hand&#x27;: 1, ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;receiving&#x27;:<br>4.162612232542636, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">assemblyline an assembly<br>line is a manufacturing ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;consider&#x27;: 1, &#x27;world&#x27;:<br>1, &#x27;bring&#x27;: 2, &#x27;pins&#x27; ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;consider&#x27;:<br>4.336965619687414, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">astronomicalunit an<br>astronomical unit ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;given&#x27;: 1, &#x27;ephemeris&#x27;:<br>2, &#x27;world&#x27;: 2, ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;given&#x27;:<br>2.5682202783138064, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">abbess an abbess latin<br>abbatissa feminine form ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;major&#x27;: 1, &#x27;abbess&#x27;:<br>10, &#x27;given&#x27;: 1, ...</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">{&#x27;major&#x27;:<br>2.356876809458607, ...</td>\n",
       "    </tr>\n",
       "</table>\n",
       "[72269 rows x 3 columns]<br/>Note: Only the head of the SFrame is printed.<br/>You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns.\n",
       "</div>"
      ],
      "text/plain": [
       "Columns:\n",
       "\tX1\tstr\n",
       "\tbow\tdict\n",
       "\ttfidf\tdict\n",
       "\n",
       "Rows: 72269\n",
       "\n",
       "Data:\n",
       "+-------------------------------+-------------------------------+\n",
       "|               X1              |              bow              |\n",
       "+-------------------------------+-------------------------------+\n",
       "| aynrand born and educated ... | {'spoke': 1, '5000': 1, 'f... |\n",
       "| asphalt in american englis... | {'lain': 1, 'commonly': 4,... |\n",
       "| actinopterygii the actinop... | {'what': 1, 'follows': 1, ... |\n",
       "| altaiclanguages these lang... | {'follows': 2, 'has': 11, ... |\n",
       "| argon the name argon is de... | {'commonly': 1, 'lattice':... |\n",
       "| augustderleth a 1938 gugge... | {'rescue': 2, 'amoral': 1,... |\n",
       "| amateur amateurism can be ... | {'receiving': 1, 'having':... |\n",
       "| assemblyline an assembly l... | {'consider': 1, 'world': 1... |\n",
       "| astronomicalunit an astron... | {'given': 1, 'ephemeris': ... |\n",
       "| abbess an abbess latin abb... | {'major': 1, 'abbess': 10,... |\n",
       "+-------------------------------+-------------------------------+\n",
       "+-------------------------------+\n",
       "|             tfidf             |\n",
       "+-------------------------------+\n",
       "| {'spoke': 4.83030828067305... |\n",
       "| {'lain': 8.480100346078947... |\n",
       "| {'what': 2.505781957805935... |\n",
       "| {'follows': 7.511333478528... |\n",
       "| {'commonly': 3.77177206798... |\n",
       "| {'rescue': 9.1902120260774... |\n",
       "| {'receiving': 4.1626122325... |\n",
       "| {'consider': 4.33696561968... |\n",
       "| {'given': 2.56822027831380... |\n",
       "| {'major': 2.35687680945860... |\n",
       "+-------------------------------+\n",
       "[72269 rows x 3 columns]\n",
       "Note: Only the head of the SFrame is printed.\n",
       "You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns."
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:12:21.040646Z",
     "start_time": "2019-06-14T15:12:21.012144Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('spoke', 4.830308280673057),\n",
       " ('5000', 4.791220891965009),\n",
       " ('follows', 3.7556667392640373),\n",
       " ('given', 2.5682202783138064),\n",
       " ('percent', 17.481279902908025)]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(sf['tfidf'][0].items())[:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Text cleaning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:12:25.655792Z",
     "start_time": "2019-06-14T15:12:25.651399Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [],
   "source": [
    "docs = sf['bow'].dict_trim_by_values(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:14:09.998427Z",
     "start_time": "2019-06-14T15:14:09.991281Z"
    },
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "outputs": [],
   "source": [
    "docs = docs.dict_trim_by_keys(\n",
    "    tc.text_analytics.stop_words(),\n",
    "    exclude=True)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Topic modeling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:14:15.739727Z",
     "start_time": "2019-06-14T15:14:15.734950Z"
    },
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Help on function create in module turicreate.toolkits.topic_model.topic_model:\n",
      "\n",
      "create(dataset, num_topics=10, initial_topics=None, alpha=None, beta=0.1, num_iterations=10, num_burnin=5, associations=None, verbose=False, print_interval=10, validation_set=None, method='auto')\n",
      "    Create a topic model from the given data set. A topic model assumes each\n",
      "    document is a mixture of a set of topics, where for each topic some words\n",
      "    are more likely than others. One statistical approach to do this is called a\n",
      "    \"topic model\". This method learns a topic model for the given document\n",
      "    collection.\n",
      "    \n",
      "    Parameters\n",
      "    ----------\n",
      "    dataset : SArray of type dict or SFrame with a single column of type dict\n",
      "        A bag of words representation of a document corpus.\n",
      "        Each element is a dictionary representing a single document, where\n",
      "        the keys are words and the values are the number of times that word\n",
      "        occurs in that document.\n",
      "    \n",
      "    num_topics : int, optional\n",
      "        The number of topics to learn.\n",
      "    \n",
      "    initial_topics : SFrame, optional\n",
      "        An SFrame with a column of unique words representing the vocabulary\n",
      "        and a column of dense vectors representing\n",
      "        probability of that word given each topic. When provided,\n",
      "        these values are used to initialize the algorithm.\n",
      "    \n",
      "    alpha : float, optional\n",
      "        Hyperparameter that controls the diversity of topics in a document.\n",
      "        Smaller values encourage fewer topics per document.\n",
      "        Provided value must be positive. Default value is 50/num_topics.\n",
      "    \n",
      "    beta : float, optional\n",
      "        Hyperparameter that controls the diversity of words in a topic.\n",
      "        Smaller values encourage fewer words per topic. Provided value\n",
      "        must be positive.\n",
      "    \n",
      "    num_iterations : int, optional\n",
      "        The number of iterations to perform.\n",
      "    \n",
      "    num_burnin : int, optional\n",
      "        The number of iterations to perform when inferring the topics for\n",
      "        documents at prediction time.\n",
      "    \n",
      "    verbose : bool, optional\n",
      "        When True, print most probable words for each topic while printing\n",
      "        progress.\n",
      "    \n",
      "    print_interval : int, optional\n",
      "        The number of iterations to wait between progress reports.\n",
      "    \n",
      "    associations : SFrame, optional\n",
      "        An SFrame with two columns named \"word\" and \"topic\" containing words\n",
      "        and the topic id that the word should be associated with. These words\n",
      "        are not considered during learning.\n",
      "    \n",
      "    validation_set : SArray of type dict or SFrame with a single column\n",
      "        A bag of words representation of a document corpus, similar to the\n",
      "        format required for `dataset`. This will be used to monitor model\n",
      "        performance during training. Each document in the provided validation\n",
      "        set is randomly split: the first portion is used estimate which topic\n",
      "        each document belongs to, and the second portion is used to estimate\n",
      "        the model's performance at predicting the unseen words in the test data.\n",
      "    \n",
      "    method : {'cgs', 'alias'}, optional\n",
      "        The algorithm used for learning the model.\n",
      "    \n",
      "        - *cgs:* Collapsed Gibbs sampling\n",
      "        - *alias:* AliasLDA method.\n",
      "    \n",
      "    Returns\n",
      "    -------\n",
      "    out : TopicModel\n",
      "        A fitted topic model. This can be used with\n",
      "        :py:func:`~TopicModel.get_topics()` and\n",
      "        :py:func:`~TopicModel.predict()`. While fitting is in progress, several\n",
      "        metrics are shown, including:\n",
      "    \n",
      "        +------------------+---------------------------------------------------+\n",
      "        |      Field       | Description                                       |\n",
      "        +==================+===================================================+\n",
      "        | Elapsed Time     | The number of elapsed seconds.                    |\n",
      "        +------------------+---------------------------------------------------+\n",
      "        | Tokens/second    | The number of unique words processed per second   |\n",
      "        +------------------+---------------------------------------------------+\n",
      "        | Est. Perplexity  | An estimate of the model's ability to model the   |\n",
      "        |                  | training data. See the documentation on evaluate. |\n",
      "        +------------------+---------------------------------------------------+\n",
      "    \n",
      "    See Also\n",
      "    --------\n",
      "    TopicModel, TopicModel.get_topics, TopicModel.predict,\n",
      "    turicreate.SArray.dict_trim_by_keys, TopicModel.evaluate\n",
      "    \n",
      "    References\n",
      "    ----------\n",
      "    - `Wikipedia - Latent Dirichlet allocation\n",
      "      <http://en.wikipedia.org/wiki/Latent_Dirichlet_allocation>`_\n",
      "    \n",
      "    - Alias method: Li, A. et al. (2014) `Reducing the Sampling Complexity of\n",
      "      Topic Models. <http://www.sravi.org/pubs/fastlda-kdd2014.pdf>`_.\n",
      "      KDD 2014.\n",
      "    \n",
      "    Examples\n",
      "    --------\n",
      "    The following example includes an SArray of documents, where\n",
      "    each element represents a document in \"bag of words\" representation\n",
      "    -- a dictionary with word keys and whose values are the number of times\n",
      "    that word occurred in the document:\n",
      "    \n",
      "    >>> docs = turicreate.SArray('https://static.turi.com/datasets/nytimes')\n",
      "    \n",
      "    Once in this form, it is straightforward to learn a topic model.\n",
      "    \n",
      "    >>> m = turicreate.topic_model.create(docs)\n",
      "    \n",
      "    It is also easy to create a new topic model from an old one  -- whether\n",
      "    it was created using Turi Create or another package.\n",
      "    \n",
      "    >>> m2 = turicreate.topic_model.create(docs, initial_topics=m['topics'])\n",
      "    \n",
      "    To manually fix several words to always be assigned to a topic, use\n",
      "    the `associations` argument. The following will ensure that topic 0\n",
      "    has the most probability for each of the provided words:\n",
      "    \n",
      "    >>> from turicreate import SFrame\n",
      "    >>> associations = SFrame({'word':['hurricane', 'wind', 'storm'],\n",
      "                               'topic': [0, 0, 0]})\n",
      "    >>> m = turicreate.topic_model.create(docs,\n",
      "                                        associations=associations)\n",
      "    \n",
      "    More advanced usage allows you  to control aspects of the model and the\n",
      "    learning method.\n",
      "    \n",
      "    >>> import turicreate as tc\n",
      "    >>> m = tc.topic_model.create(docs,\n",
      "                                  num_topics=20,       # number of topics\n",
      "                                  num_iterations=10,   # algorithm parameters\n",
      "                                  alpha=.01, beta=.1)  # hyperparameters\n",
      "    \n",
      "    To evaluate the model's ability to generalize, we can create a train/test\n",
      "    split where a portion of the words in each document are held out from\n",
      "    training.\n",
      "    \n",
      "    >>> train, test = tc.text_analytics.random_split(.8)\n",
      "    >>> m = tc.topic_model.create(train)\n",
      "    >>> results = m.evaluate(test)\n",
      "    >>> print results['perplexity']\n",
      "\n"
     ]
    }
   ],
   "source": [
    "help(tc.topic_model.create)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:14:20.794184Z",
     "start_time": "2019-06-14T15:14:20.790797Z"
    },
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Help on function random_split in module turicreate.toolkits.text_analytics._util:\n",
      "\n",
      "random_split(dataset, prob=0.5)\n",
      "    Utility for performing a random split for text data that is already in\n",
      "    bag-of-words format. For each (word, count) pair in a particular element,\n",
      "    the counts are uniformly partitioned in either a training set or a test\n",
      "    set.\n",
      "    \n",
      "    Parameters\n",
      "    ----------\n",
      "    dataset : SArray of type dict, SFrame with columns of type dict\n",
      "        A data set in bag-of-words format.\n",
      "    \n",
      "    prob : float, optional\n",
      "        Probability for sampling a word to be placed in the test set.\n",
      "    \n",
      "    Returns\n",
      "    -------\n",
      "    train, test : SArray\n",
      "        Two data sets in bag-of-words format, where the combined counts are\n",
      "        equal to the counts in the original data set.\n",
      "    \n",
      "    Examples\n",
      "    --------\n",
      "    >>> docs = turicreate.SArray([{'are':5, 'you':3, 'not': 1, 'entertained':10}])\n",
      "    >>> train, test = turicreate.text_analytics.random_split(docs)\n",
      "    >>> print(train)\n",
      "    [{'not': 1.0, 'you': 3.0, 'are': 3.0, 'entertained': 7.0}]\n",
      "    >>> print(test)\n",
      "    [{'are': 2.0, 'entertained': 3.0}]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "help(tc.text_analytics.random_split)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:14:38.091836Z",
     "start_time": "2019-06-14T15:14:24.927630Z"
    },
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "train, test = tc.text_analytics.random_split(docs, .8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:15:01.989430Z",
     "start_time": "2019-06-14T15:14:42.688161Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<pre>Learning a topic model</pre>"
      ],
      "text/plain": [
       "Learning a topic model"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>       Number of documents     72269</pre>"
      ],
      "text/plain": [
       "       Number of documents     72269"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>           Vocabulary size    108205</pre>"
      ],
      "text/plain": [
       "           Vocabulary size    108205"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>   Running collapsed Gibbs sampling</pre>"
      ],
      "text/plain": [
       "   Running collapsed Gibbs sampling"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+---------------+----------------+-----------------+</pre>"
      ],
      "text/plain": [
       "+-----------+---------------+----------------+-----------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| Iteration | Elapsed Time  | Tokens/Second  | Est. Perplexity |</pre>"
      ],
      "text/plain": [
       "| Iteration | Elapsed Time  | Tokens/Second  | Est. Perplexity |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+---------------+----------------+-----------------+</pre>"
      ],
      "text/plain": [
       "+-----------+---------------+----------------+-----------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 10        | 1.66s         | 6.17013e+06    | 0               |</pre>"
      ],
      "text/plain": [
       "| 10        | 1.66s         | 6.17013e+06    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 20        | 3.12s         | 6.57117e+06    | 0               |</pre>"
      ],
      "text/plain": [
       "| 20        | 3.12s         | 6.57117e+06    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 30        | 4.58s         | 6.41968e+06    | 0               |</pre>"
      ],
      "text/plain": [
       "| 30        | 4.58s         | 6.41968e+06    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 40        | 6.00s         | 6.61674e+06    | 0               |</pre>"
      ],
      "text/plain": [
       "| 40        | 6.00s         | 6.61674e+06    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 50        | 7.42s         | 6.53873e+06    | 0               |</pre>"
      ],
      "text/plain": [
       "| 50        | 7.42s         | 6.53873e+06    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 60        | 8.85s         | 6.46591e+06    | 0               |</pre>"
      ],
      "text/plain": [
       "| 60        | 8.85s         | 6.46591e+06    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 70        | 10.38s        | 5.92867e+06    | 0               |</pre>"
      ],
      "text/plain": [
       "| 70        | 10.38s        | 5.92867e+06    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 80        | 11.90s        | 6.36375e+06    | 0               |</pre>"
      ],
      "text/plain": [
       "| 80        | 11.90s        | 6.36375e+06    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 90        | 13.42s        | 6.20572e+06    | 0               |</pre>"
      ],
      "text/plain": [
       "| 90        | 13.42s        | 6.20572e+06    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 100       | 15.03s        | 5.4879e+06     | 0               |</pre>"
      ],
      "text/plain": [
       "| 100       | 15.03s        | 5.4879e+06     | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+---------------+----------------+-----------------+</pre>"
      ],
      "text/plain": [
       "+-----------+---------------+----------------+-----------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "m = tc.topic_model.create(train, \n",
    "                                num_topics=100,       # number of topics\n",
    "                                num_iterations=100,   # algorithm parameters\n",
    "                                alpha=None, beta=.1)  # hyperparameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:15:19.366488Z",
     "start_time": "2019-06-14T15:15:12.923625Z"
    },
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4552.105441414741\n"
     ]
    }
   ],
   "source": [
    "results = m.evaluate(test)\n",
    "print(results['perplexity'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:15:22.298503Z",
     "start_time": "2019-06-14T15:15:22.265410Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Class                          : TopicModel\n",
       "\n",
       "Schema\n",
       "------\n",
       "Vocabulary Size                : 108205\n",
       "\n",
       "Settings\n",
       "--------\n",
       "Number of Topics               : 100\n",
       "alpha                          : 0.5\n",
       "beta                           : 0.1\n",
       "Iterations                     : 100\n",
       "Training time                  : 16.0432\n",
       "Verbose                        : True\n",
       "\n",
       "Accessible fields             : \n",
       "m.topics                      : An SFrame containing the topics.\n",
       "m.vocabulary                  : An SArray containing the words in the vocabulary.\n",
       "Useful methods                : \n",
       "m.get_topics()                : Get the most probable words per topic.\n",
       "m.predict(new_docs)           : Make predictions for new documents."
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:15:26.257938Z",
     "start_time": "2019-06-14T15:15:26.096754Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\"><table frame=\"box\" rules=\"cols\">\n",
       "    <tr>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">topic</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">word</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">score</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">years</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.004647514462204498</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">evans</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.004059221492305195</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">lebanon</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0028172696669622205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">green</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0028172696669622205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">time</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0020982449259741827</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">national</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.005351237598960527</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">back</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.002278117379832135</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">baldwin</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0018772756121197358</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">chicago</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0018104686508343358</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">1</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">private</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0016100477669781365</td>\n",
       "    </tr>\n",
       "</table>\n",
       "[500 rows x 3 columns]<br/>Note: Only the head of the SFrame is printed.<br/>You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns.\n",
       "</div>"
      ],
      "text/plain": [
       "Columns:\n",
       "\ttopic\tint\n",
       "\tword\tstr\n",
       "\tscore\tfloat\n",
       "\n",
       "Rows: 500\n",
       "\n",
       "Data:\n",
       "+-------+----------+-----------------------+\n",
       "| topic |   word   |         score         |\n",
       "+-------+----------+-----------------------+\n",
       "|   0   |  years   |  0.004647514462204498 |\n",
       "|   0   |  evans   |  0.004059221492305195 |\n",
       "|   0   | lebanon  | 0.0028172696669622205 |\n",
       "|   0   |  green   | 0.0028172696669622205 |\n",
       "|   0   |   time   | 0.0020982449259741827 |\n",
       "|   1   | national |  0.005351237598960527 |\n",
       "|   1   |   back   |  0.002278117379832135 |\n",
       "|   1   | baldwin  | 0.0018772756121197358 |\n",
       "|   1   | chicago  | 0.0018104686508343358 |\n",
       "|   1   | private  | 0.0016100477669781365 |\n",
       "+-------+----------+-----------------------+\n",
       "[500 rows x 3 columns]\n",
       "Note: Only the head of the SFrame is printed.\n",
       "You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns."
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m.get_topics()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:15:30.005613Z",
     "start_time": "2019-06-14T15:15:30.001439Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Help on method get_topics in module turicreate.toolkits.topic_model.topic_model:\n",
      "\n",
      "get_topics(topic_ids=None, num_words=5, cdf_cutoff=1.0, output_type='topic_probabilities') method of turicreate.toolkits.topic_model.topic_model.TopicModel instance\n",
      "    Get the words associated with a given topic. The score column is the\n",
      "    probability of choosing that word given that you have chosen a\n",
      "    particular topic.\n",
      "    \n",
      "    Parameters\n",
      "    ----------\n",
      "    topic_ids : list of int, optional\n",
      "        The topics to retrieve words. Topic ids are zero-based.\n",
      "        Throws an error if greater than or equal to m['num_topics'], or\n",
      "        if the requested topic name is not present.\n",
      "    \n",
      "    num_words : int, optional\n",
      "        The number of words to show.\n",
      "    \n",
      "    cdf_cutoff : float, optional\n",
      "        Allows one to only show the most probable words whose cumulative\n",
      "        probability is below this cutoff. For example if there exist\n",
      "        three words where\n",
      "    \n",
      "        .. math::\n",
      "           p(word_1 | topic_k) = .1\n",
      "    \n",
      "           p(word_2 | topic_k) = .2\n",
      "    \n",
      "           p(word_3 | topic_k) = .05\n",
      "    \n",
      "        then setting :math:`cdf_{cutoff}=.3` would return only\n",
      "        :math:`word_1` and :math:`word_2` since\n",
      "        :math:`p(word_1 | topic_k) + p(word_2 | topic_k) <= cdf_{cutoff}`\n",
      "    \n",
      "    output_type : {'topic_probabilities' | 'topic_words'}, optional\n",
      "        Determine the type of desired output. See below.\n",
      "    \n",
      "    Returns\n",
      "    -------\n",
      "    out : SFrame\n",
      "        If output_type is 'topic_probabilities', then the returned value is\n",
      "        an SFrame with a column of words ranked by a column of scores for\n",
      "        each topic. Otherwise, the returned value is a SArray where\n",
      "        each element is a list of the most probable words for each topic.\n",
      "    \n",
      "    Examples\n",
      "    --------\n",
      "    Get the highest ranked words for all topics.\n",
      "    \n",
      "    >>> docs = turicreate.SArray('https://static.turi.com/datasets/nips-text')\n",
      "    >>> m = turicreate.topic_model.create(docs,\n",
      "                                        num_iterations=50)\n",
      "    >>> m.get_topics()\n",
      "    +-------+----------+-----------------+\n",
      "    | topic |   word   |      score      |\n",
      "    +-------+----------+-----------------+\n",
      "    |   0   |   cell   |  0.028974400831 |\n",
      "    |   0   |  input   | 0.0259470208503 |\n",
      "    |   0   |  image   | 0.0215721599763 |\n",
      "    |   0   |  visual  | 0.0173635081992 |\n",
      "    |   0   |  object  | 0.0172447874156 |\n",
      "    |   1   | function | 0.0482834508265 |\n",
      "    |   1   |  input   | 0.0456270024091 |\n",
      "    |   1   |  point   | 0.0302662839454 |\n",
      "    |   1   |  result  | 0.0239474934631 |\n",
      "    |   1   | problem  | 0.0231750116011 |\n",
      "    |  ...  |   ...    |       ...       |\n",
      "    +-------+----------+-----------------+\n",
      "    \n",
      "    Get the highest ranked words for topics 0 and 1 and show 15 words per\n",
      "    topic.\n",
      "    \n",
      "    >>> m.get_topics([0, 1], num_words=15)\n",
      "    +-------+----------+------------------+\n",
      "    | topic |   word   |      score       |\n",
      "    +-------+----------+------------------+\n",
      "    |   0   |   cell   |  0.028974400831  |\n",
      "    |   0   |  input   | 0.0259470208503  |\n",
      "    |   0   |  image   | 0.0215721599763  |\n",
      "    |   0   |  visual  | 0.0173635081992  |\n",
      "    |   0   |  object  | 0.0172447874156  |\n",
      "    |   0   | response | 0.0139740298286  |\n",
      "    |   0   |  layer   | 0.0122585145062  |\n",
      "    |   0   | features | 0.0115343177265  |\n",
      "    |   0   | feature  | 0.0103530459301  |\n",
      "    |   0   | spatial  | 0.00823387994361 |\n",
      "    |  ...  |   ...    |       ...        |\n",
      "    +-------+----------+------------------+\n",
      "    \n",
      "    If one wants to instead just get the top words per topic, one may\n",
      "    change the format of the output as follows.\n",
      "    \n",
      "    >>> topics = m.get_topics(output_type='topic_words')\n",
      "    dtype: list\n",
      "    Rows: 10\n",
      "    [['cell', 'image', 'input', 'object', 'visual'],\n",
      "     ['algorithm', 'data', 'learning', 'method', 'set'],\n",
      "     ['function', 'input', 'point', 'problem', 'result'],\n",
      "     ['model', 'output', 'pattern', 'set', 'unit'],\n",
      "     ['action', 'learning', 'net', 'problem', 'system'],\n",
      "     ['error', 'function', 'network', 'parameter', 'weight'],\n",
      "     ['information', 'level', 'neural', 'threshold', 'weight'],\n",
      "     ['control', 'field', 'model', 'network', 'neuron'],\n",
      "     ['hidden', 'layer', 'system', 'training', 'vector'],\n",
      "     ['component', 'distribution', 'local', 'model', 'optimal']]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "help(m.get_topics)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:15:40.870971Z",
     "start_time": "2019-06-14T15:15:40.694223Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['storm', 'florida', 'area', 'due', 'texas', 'people', 'damage', 'hurricane', 'tropical', 'system']\n",
      "['project', 'ryan', 'founded', 'harvard', 'carroll', 'including', 'wilson', 'school', 'oxford', 'national']\n",
      "['alliance', 'organization', 'membership', 'board', 'member', 'members', 'association', 'groups', 'group', 'society']\n",
      "['summer', 'george', 'kan', 'julian', '1978', 'date', 'years', 'wolfe', 'london', 'italian']\n",
      "['german', 'division', 'forces', 'british', 'army', 'men', 'battle', 'general', 'war', 'military']\n",
      "['arrow', 'williams', 'california', 'warren', 'santa', 'los', 'san', 'great', 'angeles', 'renamed']\n",
      "['miller', 'time', 'produced', 'years', 'lewis', 'hamilton', 'morris', '1999', '2005', 'epic']\n",
      "['played', 'games', 'coach', 'won', 'points', 'team', 'game', 'record', 'season', 'teams']\n",
      "['green', 'time', 'connecticut', 'evans', 'years', 'sixth', 'lebanon', '2001', 'worked', '2005']\n",
      "['van', 'berlin', 'im', 'des', 'von', 'die', 'den', 'der', 'das', 'und']\n",
      "['clothing', 'hat', 'worn', 'wear', 'made', 'green', 'summer', 'black', 'top', 'dress']\n",
      "['2009', 'year', 'kids', 'october', '2010', '2007', 'school', '2006', '2008', 'national']\n",
      "['magazine', 'grant', 'oregon', 'portland', 'college', '1985', '2006', 'year', 'long', 'beer']\n",
      "['network', 'show', 'local', 'news', 'broadcast', 'radio', 'channel', 'stations', 'television', 'station']\n",
      "['bridge', 'north', 'river', 'county', 'state', 'west', 'route', 'city', 'east', 'road']\n",
      "['part', 'degree', 'na', 'valid', 'agent', 'history', 'reform', 'born', 'agency', 'trn']\n",
      "['children', 'father', 'mother', 'married', 'years', 'family', 'born', 'life', 'died', 'time']\n",
      "['early', 'population', 'century', 'american', 'war', 'government', 'states', 'british', 'united', 'people']\n",
      "['stories', 'magazine', 'writing', 'history', 'work', 'wrote', 'works', 'published', 'book', 'books']\n",
      "['collection', 'painting', 'arts', 'museum', 'work', 'york', 'design', 'artists', 'art', 'works']\n",
      "['named', 'flag', 'orange', 'blue', 'colors', 'yellow', 'tiger', 'black', 'red', 'white']\n",
      "['made', 'mission', 'test', 'project', 'space', 'aircraft', 'wright', 'design', 'launch', 'flight']\n",
      "['alan', 'phillips', 'preov', 'koice', 'beth', 'trenn', 'phillip', 'nitra', 'town', 'trnava']\n",
      "['racing', 'cars', 'car', 'engines', 'models', 'series', 'engine', 'system', 'race', 'model']\n",
      "['building', 'duck', 'london', 'philadelphia', 'represented', 'major', 'york', 'design', 'journal', 'duncan']\n",
      "['english', 'language', 'form', 'meaning', 'word', 'words', 'names', 'written', 'languages', 'common']\n",
      "['gene', 'enzyme', 'dna', 'site', 'proteins', 'protein', 'cells', 'cell', 'structure', 'genes']\n",
      "['work', 'year', '2009', 'andrew', 'list', 'henin', 'link', 'dates', 'part', 'calendar']\n",
      "['law', 'public', 'united', 'state', 'legal', 'act', 'states', 'court', 'case', 'federal']\n",
      "['trains', 'west', 'line', 'rail', 'train', 'station', 'services', 'service', 'opened', 'railway']\n",
      "['served', 'cemetery', 'received', 'texas', 'po', 'christi', 'university', 'post', 'corpus', 'mexico']\n",
      "['international', 'worked', 'festival', 'director', 'film', 'awards', 'won', 'award', 'awarded', 'academy']\n",
      "['empire', 'china', 'chinese', 'dynasty', 'roman', 'greek', 'time', 'king', 'bc', 'emperor']\n",
      "['port', 'ship', 'coast', 'island', 'sea', 'fleet', 'region', 'bay', 'ships', 'islands']\n",
      "['south', 'north', 'river', 'mountain', 'water', 'creek', 'lake', 'valley', 'park', 'area']\n",
      "['january', 'day', 'june', '2009', 'october', 'december', '2010', '2007', '2008', 'september']\n",
      "['round', 'team', 'won', 'event', 'title', 'world', 'match', 'championship', 'lost', 'open']\n",
      "['created', '2010', 'ash', 'complex', 'sanders', 'hotel', 'skating', 'school', 'house', 'msn']\n",
      "['left', 'easy', 'leg', 'ancient', 'benson', 'school', '2003', 'critical', 'british', 'westfield']\n",
      "['released', 'world', 'games', 'version', 'series', 'video', 'player', 'game', 'players', '2']\n",
      "['won', 'cup', 'team', 'teams', 'season', 'final', 'football', 'club', 'league', 'played']\n",
      "['1988', 'called', 'list', '1971', '2010', '1994', 'made', 'trent', 'found', 'local']\n",
      "['world', 'received', 'usa', 'wall', 'part', 'start0', 'williams', 'work', 'events', 'plotdata']\n",
      "['years', 'age', 'income', 'median', '18', 'living', 'males', 'town', 'average', 'population']\n",
      "['youth', 'bamboo', 'ross', 'griffin', 'williams', 'meeting', 'monument', 'town', 'joined', 'girl']\n",
      "['south', 'european', 'international', 'economic', 'states', 'development', 'united', 'government', 'countries', 'world']\n",
      "['national', 'constituencies', 'animation', 'part', 'track', 'baldwin', '1982', 'private', 'back', 'chicago']\n",
      "['hill', 'morgan', 'davis', 'cj', '1992', 'arch', 'child', '1995', 'tony', 'part']\n",
      "['form', 'called', '1', 'theory', 'set', 'number', 'function', 'type', 'system', 'model']\n",
      "['form', 'women', 'social', 'life', 'society', 'people', 'time', 'world', 'human', 'work']\n",
      "['city', 'san', 'paris', 'el', 'france', 'la', 'de', 'french', 'spain', 'spanish']\n",
      "['city', 'located', 'house', 'park', 'building', 'street', 'built', 'town', 'area', 'century']\n",
      "['made', 'include', 'variety', 'traditional', 'rice', 'popular', 'food', 'called', 'served', 'wine']\n",
      "['canadian', 'ontario', 'alberta', 'british', 'york', '2002', 'toronto', 'quebec', 'canada', 'montreal']\n",
      "['khan', 'afghanistan', 'indian', 'singh', 'temple', 'punjab', 'pakistan', 'sri', 'india', 'community']\n",
      "['brooklyn', 'city', 'class', 'ambassador', '2004', 'due', 'babylon', 'historical', 'area', '5']\n",
      "['heat', 'iron', 'high', 'process', 'water', 'form', 'gas', 'temperature', 'nuclear', 'energy']\n",
      "['airlines', 'regional', 'airport', '2010', 'international', 'travel', 'airways', '2005', 'airline', 'joined']\n",
      "['head', '15', 'home', '2010', '2003', 'elected', 'park', 'stone', 'gaol', 'numerous']\n",
      "['poland', 'years', 'polish', 'soviet', 'moscow', 'war', 'russian', 'swedish', 'union', 'russia']\n",
      "['william', 'henry', 'england', 'sir', 'london', 'royal', 'lord', 'king', 'son', 'john']\n",
      "['ohio', 'virginia', 'district', 'washington', 'john', 'carolina', 'county', 'state', 'served', 'south']\n",
      "['state', 'election', 'political', 'minister', 'council', 'national', 'party', 'president', 'government', 'elected']\n",
      "['district', 'districts', 'register', 'properties', 'places', 'illinois', 'national', 'county', 'historic', 'school']\n",
      "['operations', 'aircraft', 'base', 'united', 'war', 'service', 'air', 'group', 'force', 'training']\n",
      "['number', '25', '2010', 'golf', 'time', 'scrooge', 'group', 'university', 'chams', 'world']\n",
      "['light', 'range', 'made', 'time', 'system', 'small', 'power', 'energy', 'high', 'current']\n",
      "['1983', 'university', 'states', 'served', 'appointed', '1986', 'degree', 'bill', 'professor', '2000']\n",
      "['years', 'head', 'joseph', 'school', 'smiths', 'rejoice', 'smith', 'clark', 'stone', 'plates']\n",
      "['found', 'occur', 'body', 'small', 'disease', 'large', 'species', 'family', 'order', 'birds']\n",
      "['number', 'fat', '2006', 'named', 'work', 'death', 'published', 'flores', 'vincent', 'villa']\n",
      "['company', 'business', 'mine', 'sold', 'oil', 'stores', 'industry', 'year', 'production', 'mining']\n",
      "['martin', 'kitty', 'obesity', 'turtle', 'point', '2000', 'camp', 'clark', 'moved', 'born']\n",
      "['1994', '1998', 'ryo', 'venezuela', 'tannins', 'national', 'group', 'minor', 'christmas', 'world']\n",
      "['parish', 'saint', 'bishop', 'church', 'roman', 'council', 'st', 'century', 'catholic', 'pope']\n",
      "['released', 'records', 'single', 'album', 'songs', 'music', 'band', 'live', 'song', 'tour']\n",
      "['wales', 'australia', 'played', 'cricket', 'zealand', 'day', 'sydney', 'made', 'australian', 'south']\n",
      "['information', 'network', 'system', 'software', 'systems', 'internet', 'users', 'technology', 'computer', 'data']\n",
      "['philippine', 'clara', 'foundation', 'chief', 'full', 'fair', 'philippines', 'manila', 'rupiah', 'belle']\n",
      "['group', 'began', 'anderson', 'minnesota', 'john', 'period', 'jones', 'te', 'td', 'dont']\n",
      "['northern', 'cork', 'dublin', 'ireland', 'irish', 'senior', 'john', 'county', 'title', 'medal']\n",
      "['school', '2006', 'stone', 'york', 'bay', 'village', 'held', 'created', 'local', 'community']\n",
      "['years', 'jersey', 'police', 'city', 'york', 'family', 'crime', 'gang', 'prison', 'arrested']\n",
      "['2001', 'american', 'saudi', 'bin', 'world', 'al', 'sh', 'including', 'laden', 'joined']\n",
      "['school', 'education', 'year', 'students', 'research', 'schools', 'college', 'high', 'university', 'program']\n",
      "['events', 'time', 'mr', 'ride', 'roller', 'part', 'including', 'bicycle', 'riders', 'bike']\n",
      "['montenegro', 'yugoslavia', 'serbia', 'school', 'serbian', 'albanian', 'albania', 'croatian', 'croatia', 'year']\n",
      "['austin', 'shells', 'place', 'peter', 'shell', 'uk', 'school', '2010', 'active', 'mark']\n",
      "['york', 'series', 'masters', 'booth', 'world', 'free', '2003', 'major', 'birmingham', 'setting']\n",
      "['role', 'appeared', 'films', 'episode', 'television', 'show', 'series', 'character', 'movie', 'film']\n",
      "['medicine', 'dr', 'training', 'care', 'hospital', 'health', 'surgery', 'center', 'dog', 'medical']\n",
      "['norwegian', 'travis', '2007', 'ste', 'bar', 'city', 'home', 'norway', 'parker', 'include']\n",
      "['jewish', 'christian', 'people', 'god', 'gods', 'religious', 'great', 'church', 'churches', 'jesus']\n",
      "['zion', 'contest', 'miss', 'relay', 'gold', 'post', 'nec', 'mori', 'yacht', 'time']\n",
      "['york', 'theatre', 'theater', 'dance', 'musical', 'orchestra', 'opera', 'music', 'performed', 'festival']\n",
      "['tax', 'money', 'financial', 'market', 'bank', 'million', 'business', 'price', 'services', 'company']\n",
      "['life', 'death', 'father', 'end', 'find', 'back', 'man', 'make', 'time', 'tells']\n",
      "['years', '2007', 'walker', 'post', '1911', 'croydon', 'burma', 'burmese', 'named', 'day']\n",
      "['high', 'vegas', 'purchased', 'alice', 'paul', 'hotel', 'scott', 'ranch', 'las', 'highlands']\n",
      "['horse', 'stakes', 'born', 'american', 'horses', 'race', 'boas', 'breed', 'english', 'racing']\n"
     ]
    }
   ],
   "source": [
    "topics = m.get_topics(num_words=10).unstack(['word','score'], \\\n",
    "                                new_column_name='topic_words')['topic_words'].apply(lambda x: x.keys())\n",
    "for topic in topics:\n",
    "    print(topic)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:16:11.030526Z",
     "start_time": "2019-06-14T15:16:11.023731Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Help on TopicModel in module turicreate.toolkits.topic_model.topic_model object:\n",
      "\n",
      "class TopicModel(turicreate.toolkits._model.Model)\n",
      " |  TopicModel objects can be used to predict the underlying topic of a\n",
      " |  document.\n",
      " |  \n",
      " |  This model cannot be constructed directly.  Instead, use\n",
      " |  :func:`turicreate.topic_model.create` to create an instance\n",
      " |  of this model. A detailed list of parameter options and code samples\n",
      " |  are available in the documentation for the create function.\n",
      " |  \n",
      " |  Method resolution order:\n",
      " |      TopicModel\n",
      " |      turicreate.toolkits._model.Model\n",
      " |      turicreate.toolkits._model.ExposeAttributesFromProxy\n",
      " |      builtins.object\n",
      " |  \n",
      " |  Methods defined here:\n",
      " |  \n",
      " |  __init__(self, model_proxy)\n",
      " |      Initialize self.  See help(type(self)) for accurate signature.\n",
      " |  \n",
      " |  __repr__(self)\n",
      " |      Print a string description of the model when the model name is entered\n",
      " |      in the terminal.\n",
      " |  \n",
      " |  __str__(self)\n",
      " |      Return a string description of the model to the ``print`` method.\n",
      " |      \n",
      " |      Returns\n",
      " |      -------\n",
      " |      out : string\n",
      " |          A description of the model.\n",
      " |  \n",
      " |  evaluate(self, train_data, test_data=None, metric='perplexity')\n",
      " |      Estimate the model's ability to predict new data. Imagine you have a\n",
      " |      corpus of books. One common approach to evaluating topic models is to\n",
      " |      train on the first half of all of the books and see how well the model\n",
      " |      predicts the second half of each book.\n",
      " |      \n",
      " |      This method returns a metric called perplexity, which  is related to the\n",
      " |      likelihood of observing these words under the given model. See\n",
      " |      :py:func:`~turicreate.topic_model.perplexity` for more details.\n",
      " |      \n",
      " |      The provided `train_data` and `test_data` must have the same length,\n",
      " |      i.e., both data sets must have the same number of documents; the model\n",
      " |      will use train_data to estimate which topic the document belongs to, and\n",
      " |      this is used to estimate the model's performance at predicting the\n",
      " |      unseen words in the test data.\n",
      " |      \n",
      " |      See :py:func:`~turicreate.topic_model.TopicModel.predict` for details\n",
      " |      on how these predictions are made, and see\n",
      " |      :py:func:`~turicreate.text_analytics.random_split` for a helper function\n",
      " |      that can be used for making train/test splits.\n",
      " |      \n",
      " |      Parameters\n",
      " |      ----------\n",
      " |      train_data : SArray or SFrame\n",
      " |          A set of documents to predict topics for.\n",
      " |      \n",
      " |      test_data : SArray or SFrame, optional\n",
      " |          A set of documents to evaluate performance on.\n",
      " |          By default this will set to be the same as train_data.\n",
      " |      \n",
      " |      metric : str\n",
      " |          The chosen metric to use for evaluating the topic model.\n",
      " |          Currently only 'perplexity' is supported.\n",
      " |      \n",
      " |      Returns\n",
      " |      -------\n",
      " |      out : dict\n",
      " |          The set of estimated evaluation metrics.\n",
      " |      \n",
      " |      See Also\n",
      " |      --------\n",
      " |      predict, turicreate.toolkits.text_analytics.random_split\n",
      " |      \n",
      " |      Examples\n",
      " |      --------\n",
      " |      >>> docs = turicreate.SArray('https://static.turi.com/datasets/nips-text')\n",
      " |      >>> train_data, test_data = turicreate.text_analytics.random_split(docs)\n",
      " |      >>> m = turicreate.topic_model.create(train_data)\n",
      " |      >>> m.evaluate(train_data, test_data)\n",
      " |      {'perplexity': 2467.530370396021}\n",
      " |  \n",
      " |  get_topics(self, topic_ids=None, num_words=5, cdf_cutoff=1.0, output_type='topic_probabilities')\n",
      " |      Get the words associated with a given topic. The score column is the\n",
      " |      probability of choosing that word given that you have chosen a\n",
      " |      particular topic.\n",
      " |      \n",
      " |      Parameters\n",
      " |      ----------\n",
      " |      topic_ids : list of int, optional\n",
      " |          The topics to retrieve words. Topic ids are zero-based.\n",
      " |          Throws an error if greater than or equal to m['num_topics'], or\n",
      " |          if the requested topic name is not present.\n",
      " |      \n",
      " |      num_words : int, optional\n",
      " |          The number of words to show.\n",
      " |      \n",
      " |      cdf_cutoff : float, optional\n",
      " |          Allows one to only show the most probable words whose cumulative\n",
      " |          probability is below this cutoff. For example if there exist\n",
      " |          three words where\n",
      " |      \n",
      " |          .. math::\n",
      " |             p(word_1 | topic_k) = .1\n",
      " |      \n",
      " |             p(word_2 | topic_k) = .2\n",
      " |      \n",
      " |             p(word_3 | topic_k) = .05\n",
      " |      \n",
      " |          then setting :math:`cdf_{cutoff}=.3` would return only\n",
      " |          :math:`word_1` and :math:`word_2` since\n",
      " |          :math:`p(word_1 | topic_k) + p(word_2 | topic_k) <= cdf_{cutoff}`\n",
      " |      \n",
      " |      output_type : {'topic_probabilities' | 'topic_words'}, optional\n",
      " |          Determine the type of desired output. See below.\n",
      " |      \n",
      " |      Returns\n",
      " |      -------\n",
      " |      out : SFrame\n",
      " |          If output_type is 'topic_probabilities', then the returned value is\n",
      " |          an SFrame with a column of words ranked by a column of scores for\n",
      " |          each topic. Otherwise, the returned value is a SArray where\n",
      " |          each element is a list of the most probable words for each topic.\n",
      " |      \n",
      " |      Examples\n",
      " |      --------\n",
      " |      Get the highest ranked words for all topics.\n",
      " |      \n",
      " |      >>> docs = turicreate.SArray('https://static.turi.com/datasets/nips-text')\n",
      " |      >>> m = turicreate.topic_model.create(docs,\n",
      " |                                          num_iterations=50)\n",
      " |      >>> m.get_topics()\n",
      " |      +-------+----------+-----------------+\n",
      " |      | topic |   word   |      score      |\n",
      " |      +-------+----------+-----------------+\n",
      " |      |   0   |   cell   |  0.028974400831 |\n",
      " |      |   0   |  input   | 0.0259470208503 |\n",
      " |      |   0   |  image   | 0.0215721599763 |\n",
      " |      |   0   |  visual  | 0.0173635081992 |\n",
      " |      |   0   |  object  | 0.0172447874156 |\n",
      " |      |   1   | function | 0.0482834508265 |\n",
      " |      |   1   |  input   | 0.0456270024091 |\n",
      " |      |   1   |  point   | 0.0302662839454 |\n",
      " |      |   1   |  result  | 0.0239474934631 |\n",
      " |      |   1   | problem  | 0.0231750116011 |\n",
      " |      |  ...  |   ...    |       ...       |\n",
      " |      +-------+----------+-----------------+\n",
      " |      \n",
      " |      Get the highest ranked words for topics 0 and 1 and show 15 words per\n",
      " |      topic.\n",
      " |      \n",
      " |      >>> m.get_topics([0, 1], num_words=15)\n",
      " |      +-------+----------+------------------+\n",
      " |      | topic |   word   |      score       |\n",
      " |      +-------+----------+------------------+\n",
      " |      |   0   |   cell   |  0.028974400831  |\n",
      " |      |   0   |  input   | 0.0259470208503  |\n",
      " |      |   0   |  image   | 0.0215721599763  |\n",
      " |      |   0   |  visual  | 0.0173635081992  |\n",
      " |      |   0   |  object  | 0.0172447874156  |\n",
      " |      |   0   | response | 0.0139740298286  |\n",
      " |      |   0   |  layer   | 0.0122585145062  |\n",
      " |      |   0   | features | 0.0115343177265  |\n",
      " |      |   0   | feature  | 0.0103530459301  |\n",
      " |      |   0   | spatial  | 0.00823387994361 |\n",
      " |      |  ...  |   ...    |       ...        |\n",
      " |      +-------+----------+------------------+\n",
      " |      \n",
      " |      If one wants to instead just get the top words per topic, one may\n",
      " |      change the format of the output as follows.\n",
      " |      \n",
      " |      >>> topics = m.get_topics(output_type='topic_words')\n",
      " |      dtype: list\n",
      " |      Rows: 10\n",
      " |      [['cell', 'image', 'input', 'object', 'visual'],\n",
      " |       ['algorithm', 'data', 'learning', 'method', 'set'],\n",
      " |       ['function', 'input', 'point', 'problem', 'result'],\n",
      " |       ['model', 'output', 'pattern', 'set', 'unit'],\n",
      " |       ['action', 'learning', 'net', 'problem', 'system'],\n",
      " |       ['error', 'function', 'network', 'parameter', 'weight'],\n",
      " |       ['information', 'level', 'neural', 'threshold', 'weight'],\n",
      " |       ['control', 'field', 'model', 'network', 'neuron'],\n",
      " |       ['hidden', 'layer', 'system', 'training', 'vector'],\n",
      " |       ['component', 'distribution', 'local', 'model', 'optimal']]\n",
      " |  \n",
      " |  predict(self, dataset, output_type='assignment', num_burnin=None)\n",
      " |      Use the model to predict topics for each document. The provided\n",
      " |      `dataset` should be an SArray object where each element is a dict\n",
      " |      representing a single document in bag-of-words format, where keys\n",
      " |      are words and values are their corresponding counts. If `dataset` is\n",
      " |      an SFrame, then it must contain a single column of dict type.\n",
      " |      \n",
      " |      The current implementation will make inferences about each document\n",
      " |      given its estimates of the topics learned when creating the model.\n",
      " |      This is done via Gibbs sampling.\n",
      " |      \n",
      " |      Parameters\n",
      " |      ----------\n",
      " |      dataset : SArray, SFrame of type dict\n",
      " |          A set of documents to use for making predictions.\n",
      " |      \n",
      " |      output_type : str, optional\n",
      " |          The type of output desired. This can either be\n",
      " |      \n",
      " |          - assignment: the returned values are integers in [0, num_topics)\n",
      " |          - probability: each returned prediction is a vector with length\n",
      " |            num_topics, where element k represents the probability that\n",
      " |            document belongs to topic k.\n",
      " |      \n",
      " |      num_burnin : int, optional\n",
      " |          The number of iterations of Gibbs sampling to perform when\n",
      " |          inferring the topics for documents at prediction time.\n",
      " |          If provided this will override the burnin value set during\n",
      " |          training.\n",
      " |      \n",
      " |      Returns\n",
      " |      -------\n",
      " |      out : SArray\n",
      " |      \n",
      " |      See Also\n",
      " |      --------\n",
      " |      evaluate\n",
      " |      \n",
      " |      Examples\n",
      " |      --------\n",
      " |      Make predictions about which topic each document belongs to.\n",
      " |      \n",
      " |      >>> docs = turicreate.SArray('https://static.turi.com/datasets/nips-text')\n",
      " |      >>> m = turicreate.topic_model.create(docs)\n",
      " |      >>> pred = m.predict(docs)\n",
      " |      \n",
      " |      If one is interested in the probability of each topic\n",
      " |      \n",
      " |      >>> pred = m.predict(docs, output_type='probability')\n",
      " |      \n",
      " |      Notes\n",
      " |      -----\n",
      " |      For each unique word w in a document d, we sample an assignment to\n",
      " |      topic k with probability proportional to\n",
      " |      \n",
      " |      .. math::\n",
      " |          p(z_{dw} = k) \\propto (n_{d,k} + \\alpha) * \\Phi_{w,k}\n",
      " |      \n",
      " |      where\n",
      " |      \n",
      " |      - :math:`W` is the size of the vocabulary,\n",
      " |      - :math:`n_{d,k}` is the number of other times we have assigned a word in\n",
      " |        document to d to topic :math:`k`,\n",
      " |      - :math:`\\Phi_{w,k}` is the probability under the model of choosing word\n",
      " |        :math:`w` given the word is of topic :math:`k`. This is the matrix\n",
      " |        returned by calling `m['topics']`.\n",
      " |      \n",
      " |      This represents a collapsed Gibbs sampler for the document assignments\n",
      " |      while we keep the topics learned during training fixed.\n",
      " |      This process is done in parallel across all documents, five times per\n",
      " |      document.\n",
      " |  \n",
      " |  ----------------------------------------------------------------------\n",
      " |  Methods inherited from turicreate.toolkits._model.Model:\n",
      " |  \n",
      " |  save(self, location)\n",
      " |      Save the model. The model is saved as a directory which can then be\n",
      " |      loaded using the :py:func:`~turicreate.load_model` method.\n",
      " |      \n",
      " |      Parameters\n",
      " |      ----------\n",
      " |      location : string\n",
      " |          Target destination for the model. Can be a local path or remote URL.\n",
      " |      \n",
      " |      See Also\n",
      " |      ----------\n",
      " |      turicreate.load_model\n",
      " |      \n",
      " |      Examples\n",
      " |      ----------\n",
      " |      >>> model.save('my_model_file')\n",
      " |      >>> loaded_model = turicreate.load_model('my_model_file')\n",
      " |  \n",
      " |  summary(self, output=None)\n",
      " |      Print a summary of the model. The summary includes a description of\n",
      " |      training data, options, hyper-parameters, and statistics measured\n",
      " |      during model creation.\n",
      " |      \n",
      " |      Parameters\n",
      " |      ----------\n",
      " |      output : str, None\n",
      " |          The type of summary to return.\n",
      " |      \n",
      " |          - None or 'stdout' : print directly to stdout.\n",
      " |      \n",
      " |          - 'str' : string of summary\n",
      " |      \n",
      " |          - 'dict' : a dict with 'sections' and 'section_titles' ordered\n",
      " |            lists. The entries in the 'sections' list are tuples of the form\n",
      " |            ('label', 'value').\n",
      " |      \n",
      " |      Examples\n",
      " |      --------\n",
      " |      >>> m.summary()\n",
      " |  \n",
      " |  ----------------------------------------------------------------------\n",
      " |  Methods inherited from turicreate.toolkits._model.ExposeAttributesFromProxy:\n",
      " |  \n",
      " |  __dir__(self)\n",
      " |      Combine the results of dir from the current class with the results of\n",
      " |      list_fields().\n",
      " |  \n",
      " |  __getattribute__(self, attr)\n",
      " |      Use the internal proxy object for obtaining list_fields.\n",
      " |  \n",
      " |  ----------------------------------------------------------------------\n",
      " |  Data descriptors inherited from turicreate.toolkits._model.ExposeAttributesFromProxy:\n",
      " |  \n",
      " |  __dict__\n",
      " |      dictionary for instance variables (if defined)\n",
      " |  \n",
      " |  __weakref__\n",
      " |      list of weak references to the object (if defined)\n",
      " |  \n",
      " |  ----------------------------------------------------------------------\n",
      " |  Data and other attributes inherited from turicreate.toolkits._model.ExposeAttributesFromProxy:\n",
      " |  \n",
      " |  __proxy__ = None\n",
      "\n"
     ]
    }
   ],
   "source": [
    "help(m)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:16:53.661546Z",
     "start_time": "2019-06-14T15:16:53.495319Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['people', 'texas', 'tropical', 'florida', 'storm']\n",
      "['school', 'founded', 'wilson', 'harvard', 'including']\n",
      "['members', 'association', 'society', 'member', 'group']\n",
      "['summer', 'date', 'julian', 'years', 'george']\n",
      "['war', 'general', 'battle', 'german', 'army']\n",
      "['california', 'angeles', 'san', 'los', 'santa']\n",
      "['miller', 'morris', '2005', '1999', 'time']\n",
      "['game', 'season', 'team', 'points', 'games']\n",
      "['evans', 'years', 'time', 'green', 'lebanon']\n",
      "['und', 'der', 'die', 'van', 'von']\n",
      "['wear', 'worn', 'made', 'green', 'top']\n",
      "['2008', 'year', '2007', '2009', '2010']\n",
      "['oregon', 'magazine', 'year', 'portland', 'beer']\n",
      "['network', 'news', 'radio', 'show', 'station']\n",
      "['bridge', 'north', 'route', 'west', 'road']\n",
      "['part', 'born', 'na', 'agent', 'valid']\n",
      "['years', 'family', 'father', 'died', 'time']\n",
      "['united', 'population', 'people', 'american', 'states']\n",
      "['work', 'magazine', 'published', 'book', 'books']\n",
      "['collection', 'museum', 'arts', 'art', 'work']\n",
      "['white', 'black', 'red', 'flag', 'blue']\n",
      "['aircraft', 'design', 'space', 'project', 'flight']\n",
      "['trenn', 'phillip', 'nitra', 'koice', 'preov']\n",
      "['engine', 'cars', 'car', 'race', 'model']\n",
      "['london', 'duck', 'journal', 'philadelphia', 'york']\n",
      "['english', 'word', 'language', 'languages', 'words']\n",
      "['dna', 'cell', 'cells', 'enzyme', 'protein']\n",
      "['2009', 'calendar', 'year', 'link', 'list']\n",
      "['court', 'act', 'state', 'states', 'law']\n",
      "['station', 'services', 'railway', 'line', 'service']\n",
      "['cemetery', 'university', 'mexico', 'texas', 'corpus']\n",
      "['film', 'worked', 'festival', 'award', 'awards']\n",
      "['chinese', 'emperor', 'bc', 'king', 'empire']\n",
      "['islands', 'region', 'ships', 'island', 'ship']\n",
      "['water', 'park', 'lake', 'area', 'river']\n",
      "['2007', '2009', 'december', '2010', '2008']\n",
      "['match', 'world', 'title', 'won', 'championship']\n",
      "['msn', 'house', '2010', 'skating', 'school']\n",
      "['school', 'westfield', 'easy', 'british', '2003']\n",
      "['player', 'players', 'game', 'games', 'series']\n",
      "['team', 'league', 'season', 'club', 'football']\n",
      "['local', 'list', 'found', 'called', '2010']\n",
      "['plotdata', 'received', 'world', 'part', 'usa']\n",
      "['income', 'population', '18', 'years', 'age']\n",
      "['youth', 'ross', 'griffin', 'joined', 'williams']\n",
      "['south', 'government', 'international', 'countries', 'world']\n",
      "['national', 'private', 'back', 'chicago', 'baldwin']\n",
      "['cj', 'hill', 'morgan', 'davis', 'child']\n",
      "['set', '1', 'model', 'theory', 'number']\n",
      "['social', 'people', 'time', 'form', 'work']\n",
      "['french', 'de', 'spanish', 'france', 'la']\n",
      "['area', 'city', 'building', 'town', 'built']\n",
      "['food', 'popular', 'made', 'called', 'wine']\n",
      "['canadian', 'british', 'ontario', 'canada', 'toronto']\n",
      "['india', 'temple', 'khan', 'pakistan', 'indian']\n",
      "['babylon', '2004', 'area', 'class', '5']\n",
      "['water', 'process', 'nuclear', 'energy', 'form']\n",
      "['international', '2010', 'airport', 'airlines', 'airline']\n",
      "['15', 'park', 'head', 'stone', '2003']\n",
      "['soviet', 'russia', 'poland', 'russian', 'polish']\n",
      "['john', 'royal', 'william', 'sir', 'king']\n",
      "['virginia', 'county', 'served', 'state', 'carolina']\n",
      "['election', 'government', 'president', 'state', 'party']\n",
      "['school', 'county', 'national', 'districts', 'district']\n",
      "['group', 'war', 'force', 'air', 'aircraft']\n",
      "['world', 'university', '2010', 'chams', 'number']\n",
      "['energy', 'power', 'system', 'time', 'light']\n",
      "['bill', 'professor', 'degree', 'served', 'university']\n",
      "['joseph', 'head', 'smiths', 'smith', 'school']\n",
      "['birds', 'family', 'small', 'species', 'found']\n",
      "['fat', '2006', 'named', 'work', 'vincent']\n",
      "['business', 'stores', 'oil', 'company', 'mine']\n",
      "['2000', 'kitty', 'camp', 'moved', 'obesity']\n",
      "['1994', 'christmas', 'world', 'national', 'minor']\n",
      "['bishop', 'saint', 'church', 'st', 'century']\n",
      "['band', 'released', 'album', 'music', 'song']\n",
      "['made', 'zealand', 'australian', 'australia', 'south']\n",
      "['system', 'software', 'data', 'systems', 'information']\n",
      "['full', 'fair', 'philippines', 'manila', 'philippine']\n",
      "['began', 'jones', 'minnesota', 'anderson', 'td']\n",
      "['county', 'ireland', 'medal', 'irish', 'dublin']\n",
      "['school', 'held', 'village', 'local', 'community']\n",
      "['family', 'city', 'prison', 'police', 'york']\n",
      "['al', 'bin', '2001', 'joined', 'world']\n",
      "['college', 'university', 'education', 'students', 'school']\n",
      "['including', 'riders', 'ride', 'time', 'part']\n",
      "['year', 'yugoslavia', 'serbian', 'albanian', 'serbia']\n",
      "['shells', 'uk', 'school', 'austin', 'shell']\n",
      "['world', 'birmingham', 'york', 'booth', 'free']\n",
      "['show', 'role', 'episode', 'series', 'film']\n",
      "['dr', 'care', 'health', 'medical', 'hospital']\n",
      "['include', '2007', 'norway', 'norwegian', 'city']\n",
      "['religious', 'christian', 'church', 'god', 'jesus']\n",
      "['nec', 'mori', 'miss', 'post', 'time']\n",
      "['festival', 'theatre', 'dance', 'music', 'opera']\n",
      "['business', 'company', 'financial', 'bank', 'million']\n",
      "['man', 'tells', 'life', 'back', 'time']\n",
      "['years', '2007', 'post', 'croydon', 'day']\n",
      "['las', 'scott', 'vegas', 'ranch', 'hotel']\n",
      "['american', 'horses', 'horse', 'breed', 'stakes']\n"
     ]
    }
   ],
   "source": [
    "def print_topics(m):\n",
    "    topics = m.get_topics(num_words=5)\n",
    "    topics = topics.unstack(['word','score'], new_column_name='topic_words')['topic_words']\n",
    "    topics = topics.apply(lambda x: x.keys())\n",
    "    for topic in topics:\n",
    "        print(topic)\n",
    "print_topics(m)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:17:00.220909Z",
     "start_time": "2019-06-14T15:17:00.207180Z"
    },
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "> pred = m.predict(another_data) \n",
    "\n",
    "> pred = m.predict(another_data, output_type='probabilities')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Initializing from other models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:19:58.593941Z",
     "start_time": "2019-06-14T15:19:58.588723Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['__class__',\n",
       " '__delattr__',\n",
       " '__dict__',\n",
       " '__dir__',\n",
       " '__doc__',\n",
       " '__eq__',\n",
       " '__format__',\n",
       " '__ge__',\n",
       " '__getattribute__',\n",
       " '__gt__',\n",
       " '__hash__',\n",
       " '__init__',\n",
       " '__le__',\n",
       " '__lt__',\n",
       " '__module__',\n",
       " '__ne__',\n",
       " '__new__',\n",
       " '__proxy__',\n",
       " '__reduce__',\n",
       " '__reduce_ex__',\n",
       " '__repr__',\n",
       " '__setattr__',\n",
       " '__sizeof__',\n",
       " '__str__',\n",
       " '__subclasshook__',\n",
       " '__weakref__',\n",
       " '_get',\n",
       " '_get_queryable_methods',\n",
       " '_get_summary_struct',\n",
       " '_list_fields',\n",
       " '_name',\n",
       " '_native_name',\n",
       " '_training_stats',\n",
       " 'alpha',\n",
       " 'beta',\n",
       " 'evaluate',\n",
       " 'get_topics',\n",
       " 'num_burnin',\n",
       " 'num_iterations',\n",
       " 'num_topics',\n",
       " 'predict',\n",
       " 'print_interval',\n",
       " 'save',\n",
       " 'summary',\n",
       " 'topics',\n",
       " 'training_iterations',\n",
       " 'training_time',\n",
       " 'validation_time',\n",
       " 'verbose',\n",
       " 'vocabulary']"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dir(m)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:20:38.229839Z",
     "start_time": "2019-06-14T15:20:38.200206Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dtype: str\n",
       "Rows: 108205\n",
       "['book', 'political', 'time', 'readers', 'individual', 'appeared', 'peikoff', 'concepts', '100', 'picture', 'america', 'reviewers', 'philosopher', 'screenwriter', 'work', 'traditional', 'purged', 'ayn', 'york', 'articles', 'pharmacy', 'scholarship', '2001', 'designed', 'permission', 'taking', 'historian', 'library', 'russian', 'extent', 'childhood', 'respondents', 'language', 'alisa', 'writing', 'union', 'libertarian', 'positive', 'jennifer', 'notes', 'line', 'burns', 'state', 'crimea', 'sciabarra', 'based', 'rights', 'life', 'shes', 'argument', 'nonfiction', 'rejection', 'allowed', 'reason', 'culture', 'closest', 'shrugged', 'free', 'january', 'success', 'living', 'robert', 'literary', 'animated', 'american', 'reviews', 'paterson', 'people', 'percent', 'house', 'academic', 'sacrificing', 'referred', 'broadway', 'fountainhead', 'lectures', 'john', 'inspiration', 'conditions', 'lifetime', 'written', '1938', 'established', 'barbara', 'twelve', 'modern', 'final', 'intellectual', 'audience', 'stated', 'selfinterest', 'achievement', 'century', 'relationship', 'allowing', 'delivering', 'writers', 'influence', 'branden', 'film', ... ]"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m.vocabulary"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:21:11.640706Z",
     "start_time": "2019-06-14T15:21:11.272036Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "108205"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m.topics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:21:44.842823Z",
     "start_time": "2019-06-14T15:21:39.593564Z"
    },
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<pre>Initializing from provided topics and vocabulary.</pre>"
      ],
      "text/plain": [
       "Initializing from provided topics and vocabulary."
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>Learning a topic model</pre>"
      ],
      "text/plain": [
       "Learning a topic model"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>       Number of documents     72269</pre>"
      ],
      "text/plain": [
       "       Number of documents     72269"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>           Vocabulary size    171005</pre>"
      ],
      "text/plain": [
       "           Vocabulary size    171005"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>   Running collapsed Gibbs sampling</pre>"
      ],
      "text/plain": [
       "   Running collapsed Gibbs sampling"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+---------------+----------------+-----------------+</pre>"
      ],
      "text/plain": [
       "+-----------+---------------+----------------+-----------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| Iteration | Elapsed Time  | Tokens/Second  | Est. Perplexity |</pre>"
      ],
      "text/plain": [
       "| Iteration | Elapsed Time  | Tokens/Second  | Est. Perplexity |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+---------------+----------------+-----------------+</pre>"
      ],
      "text/plain": [
       "+-----------+---------------+----------------+-----------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 10        | 2.99s         | 7.251e+06      | 0               |</pre>"
      ],
      "text/plain": [
       "| 10        | 2.99s         | 7.251e+06      | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+---------------+----------------+-----------------+</pre>"
      ],
      "text/plain": [
       "+-----------+---------------+----------------+-----------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "m2 = tc.topic_model.create(docs,\n",
    "                                 num_topics=100,\n",
    "                                 initial_topics=m.topics)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "### Seeding the model with prior knowledge"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:21:53.743973Z",
     "start_time": "2019-06-14T15:21:53.739544Z"
    },
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [],
   "source": [
    "associations = tc.SFrame()\n",
    "associations['word'] = ['recognition']\n",
    "associations['topic'] = [0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:22:10.912220Z",
     "start_time": "2019-06-14T15:22:00.395606Z"
    },
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<pre>Learning a topic model</pre>"
      ],
      "text/plain": [
       "Learning a topic model"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>       Number of documents     72269</pre>"
      ],
      "text/plain": [
       "       Number of documents     72269"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>           Vocabulary size    171005</pre>"
      ],
      "text/plain": [
       "           Vocabulary size    171005"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>   Running collapsed Gibbs sampling</pre>"
      ],
      "text/plain": [
       "   Running collapsed Gibbs sampling"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+---------------+----------------+-----------------+</pre>"
      ],
      "text/plain": [
       "+-----------+---------------+----------------+-----------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| Iteration | Elapsed Time  | Tokens/Second  | Est. Perplexity |</pre>"
      ],
      "text/plain": [
       "| Iteration | Elapsed Time  | Tokens/Second  | Est. Perplexity |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+---------------+----------------+-----------------+</pre>"
      ],
      "text/plain": [
       "+-----------+---------------+----------------+-----------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 10        | 2.10s         | 1.04058e+07    | 0               |</pre>"
      ],
      "text/plain": [
       "| 10        | 2.10s         | 1.04058e+07    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 20        | 3.97s         | 1.09325e+07    | 0               |</pre>"
      ],
      "text/plain": [
       "| 20        | 3.97s         | 1.09325e+07    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 30        | 5.79s         | 9.42067e+06    | 0               |</pre>"
      ],
      "text/plain": [
       "| 30        | 5.79s         | 9.42067e+06    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 40        | 7.66s         | 1.0637e+07     | 0               |</pre>"
      ],
      "text/plain": [
       "| 40        | 7.66s         | 1.0637e+07     | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>| 50        | 9.51s         | 9.86708e+06    | 0               |</pre>"
      ],
      "text/plain": [
       "| 50        | 9.51s         | 9.86708e+06    | 0               |"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<pre>+-----------+---------------+----------------+-----------------+</pre>"
      ],
      "text/plain": [
       "+-----------+---------------+----------------+-----------------+"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "m2 = tc.topic_model.create(docs,\n",
    "                                 num_topics=20,\n",
    "                                 num_iterations=50,\n",
    "                                 associations=associations, \n",
    "                                 verbose=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:22:26.128839Z",
     "start_time": "2019-06-14T15:22:26.065521Z"
    },
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div style=\"max-height:1000px;max-width:1500px;overflow:auto;\"><table frame=\"box\" rules=\"cols\">\n",
       "    <tr>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">topic</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">word</th>\n",
       "        <th style=\"padding-left: 1em; padding-right: 1em; text-align: center\">score</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">line</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.0109206038501584</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">german</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.010542910113799127</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">de</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.010148794910641626</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">railway</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.010079824750089063</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">english</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.009242329943379372</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">chinese</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.008900763433976205</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">language</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.008654441432002766</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">china</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.008490226764020474</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">large</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.008004151346792889</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">russian</td>\n",
       "        <td style=\"padding-left: 1em; padding-right: 1em; text-align: center; vertical-align: top\">0.007705280651065117</td>\n",
       "    </tr>\n",
       "</table>\n",
       "[200 rows x 3 columns]<br/>Note: Only the head of the SFrame is printed.<br/>You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns.\n",
       "</div>"
      ],
      "text/plain": [
       "Columns:\n",
       "\ttopic\tint\n",
       "\tword\tstr\n",
       "\tscore\tfloat\n",
       "\n",
       "Rows: 200\n",
       "\n",
       "Data:\n",
       "+-------+----------+----------------------+\n",
       "| topic |   word   |        score         |\n",
       "+-------+----------+----------------------+\n",
       "|   0   |   line   |  0.0109206038501584  |\n",
       "|   0   |  german  | 0.010542910113799127 |\n",
       "|   0   |    de    | 0.010148794910641626 |\n",
       "|   0   | railway  | 0.010079824750089063 |\n",
       "|   0   | english  | 0.009242329943379372 |\n",
       "|   0   | chinese  | 0.008900763433976205 |\n",
       "|   0   | language | 0.008654441432002766 |\n",
       "|   0   |  china   | 0.008490226764020474 |\n",
       "|   0   |  large   | 0.008004151346792889 |\n",
       "|   0   | russian  | 0.007705280651065117 |\n",
       "+-------+----------+----------------------+\n",
       "[200 rows x 3 columns]\n",
       "Note: Only the head of the SFrame is printed.\n",
       "You can use print_rows(num_rows=m, num_columns=n) to print more rows and columns."
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "m2.get_topics(num_words=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-06-14T15:22:27.872285Z",
     "start_time": "2019-06-14T15:22:27.805060Z"
    },
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['german', 'line', 'english', 'de', 'railway']\n",
      "['son', 'time', 'book', 'life', 'john']\n",
      "['role', 'episode', 'show', 'film', 'series']\n",
      "['york', 'station', 'park', 'de', 'company']\n",
      "['information', 'law', 'time', 'work', 'social']\n",
      "['west', 'county', 'north', 'city', 'east']\n",
      "['years', '18', 'population', 'town', 'age']\n",
      "['games', 'team', 'game', 'won', 'season']\n",
      "['company', 'aircraft', 'air', 'force', 'division']\n",
      "['india', 'art', 'century', 'roman', 'church']\n",
      "['government', 'students', 'national', 'state', 'party']\n",
      "['schools', 'college', 'university', 'school', 'high']\n",
      "['series', 'world', 'back', 'king', 'time']\n",
      "['services', 'system', 'service', 'million', 'engine']\n",
      "['area', 'built', 'river', 'road', 'region']\n",
      "['systems', 'set', 'number', 'system', 'data']\n",
      "['water', 'small', 'species', 'food', 'found']\n",
      "['league', 'year', 'club', 'song', 'time']\n",
      "['released', 'music', 'songs', 'album', 'band']\n",
      "['army', 'united', 'states', 'court', 'war']\n"
     ]
    }
   ],
   "source": [
    "print_topics(m2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# 阅读材料\n",
    "\n",
    "<del>https://dato.com/learn/userguide/text/topic-models.html</del>\n",
    "\n",
    "https://apple.github.io/turicreate/docs/userguide/text/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "celltoolbar": "Slideshow",
  "kernelspec": {
   "display_name": "Python [conda env:anaconda]",
   "language": "python",
   "name": "conda-env-anaconda-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.4"
  },
  "latex_envs": {
   "LaTeX_envs_menu_present": true,
   "autoclose": false,
   "autocomplete": true,
   "bibliofile": "biblio.bib",
   "cite_by": "apalike",
   "current_citInitial": 1,
   "eqLabelWithNumbers": true,
   "eqNumInitial": 0,
   "hotkeys": {
    "equation": "Ctrl-E",
    "itemize": "Ctrl-I"
   },
   "labels_anchors": false,
   "latex_user_defs": false,
   "report_style_numbering": false,
   "user_envs_cfg": false
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": false,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}