{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\"セキュリティ ソフト 種類 印象 どれ No どれ 俺 ここ 種類 ウイルス ウイルス 自分 知識 稲荷 ウイルス 逆 新種 ウイルス パターン ウイルス ウイルス 新種 ウイルス 両方 イラスト ソフト 動画 ソフト 動画 ソフト テロップ 動画 作品 彼女 ポケット スマ ホ メガネ デバイス ッーン 稲荷 俺 なん それ それ 世界 メガネ デバイス 鼻 栓 デバイス 鼻 そこ 光 AR これ 彼女 彼女 立体 俺 彼女 セキュリティ ソフト ノリ クマ キャクター それ 人気 本当 俺 これ 通り 答え 誰 考え 相手 人工 知能 相手 電気 電子 機器 ローン 最新 コンピュータ 性能 ローン 中学生 社会 国会 議事堂 俺 それ 技術 家 PC ソフトウェア これ ウイルス ウェブサイト ソース コード 世界中 俺 胸 ソース コード 俺 人 人 悪意 大学 ひよっこ ソース コード 彼女 僕 それ それ 方向 兆候 読み これ 闇 稲荷 俺 バイト そっ 稲荷 自転車 颯爽 稲荷 なに ドア 鍵 母さん 鍋 声 夕飯 物 なに アンタ 感じ お袋 身の回り 稲荷 俺 調子 鬱々 サンキュー モツ 大盛り モツ やりとり あと 眠\""
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lines = open(\"./dataset/shosetsu1-wakati.txt\", \"r\") do fp\n",
    "    readline(fp)\n",
    "end"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "102-element Array{SubString{String},1}:\n",
       " \"ウェブサイト\"\n",
       " \"母さん\"   \n",
       " \"テロップ\"  \n",
       " \"大盛り\"   \n",
       " \"社会\"    \n",
       " \"イラスト\"  \n",
       " \"中学生\"   \n",
       " \"家\"     \n",
       " \"読み\"    \n",
       " \"夕飯\"    \n",
       " \"知識\"    \n",
       " \"闇\"     \n",
       " \"答え\"    \n",
       " ⋮       \n",
       " \"No\"    \n",
       " \"胸\"     \n",
       " \"方向\"    \n",
       " \"ソフト\"   \n",
       " \"颯爽\"    \n",
       " \"動画\"    \n",
       " \"ソース\"   \n",
       " \"これ\"    \n",
       " \"ひよっこ\"  \n",
       " \"鍵\"     \n",
       " \"最新\"    \n",
       " \"機器\"    "
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 重複をなくすためSet()でくくり配列にもどす．\n",
    "word = [v for v in Set(split(lines, \" \"))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "6-element Array{SubString{String},1}:\n",
       " \"programmer\"\n",
       " \"I\"         \n",
       " \"am\"        \n",
       " \"a\"         \n",
       " \"like\"      \n",
       " \"dog\"       "
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 重複をなくすためSet()でくくり配列にもどす．\n",
    "sample_text = [v for v in Set(split(\"I am a programmer like dog\", \" \"))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "sigmoid (generic function with 1 method)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "function sigmoid(z)\n",
    "    return 1 / (1 + exp(-z))\n",
    "end"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "softmax (generic function with 1 method)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "function softmax(vtw, vw)\n",
    "    sum_vtw = sum(vtw)\n",
    "    result = []\n",
    "    for i in 1:length(vtw)\n",
    "        push!(result, (exp(dot(vtw[i], vw[i])) / exp(dot(sum_vtw, vw[i]))))\n",
    "    end\n",
    "    return result\n",
    "end"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3-element Array{Float64,1}:\n",
       " 0.2\n",
       " 2.2\n",
       " 0.4"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vtw = [0.1,0.4,3.3]\n",
    "vw = [0.2,2.2,0.4]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3-element Array{Any,1}:\n",
       " 0.477114   \n",
       " 0.000564257\n",
       " 0.818731   "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "softmax(vtw',vw)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "one_hot_vectorization (generic function with 1 method)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "function one_hot_vectorization(word_array)\n",
    "    one_hot_vec = Any[]\n",
    "    for i in 1:length(word_array)\n",
    "        zeros_array = zeros(length(word_array))'\n",
    "        zeros_array[i] = 1.0\n",
    "        push!(one_hot_vec, zeros_array)\n",
    "    end\n",
    "    return one_hot_vec\n",
    "end"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "6-element Array{SubString{String},1}:\n",
       " \"programmer\"\n",
       " \"I\"         \n",
       " \"am\"        \n",
       " \"a\"         \n",
       " \"like\"      \n",
       " \"dog\"       "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample_text"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ベクトルとテキストの辞書を作る"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "vec_dict = Dict{String, Any}()\n",
    "W = one_hot_vectorization(sample_text)\n",
    "for i in 1:length(W)\n",
    "    vec_dict[sample_text[i]] = W[i]\n",
    "end"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Base.KeyIterator for a Dict{String,Any} with 6 entries. Keys:\n",
       "  \"programmer\"\n",
       "  \"I\"\n",
       "  \"am\"\n",
       "  \"a\"\n",
       "  \"like\"\n",
       "  \"dog\""
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "keys(vec_dict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Base.ValueIterator for a Dict{String,Any} with 6 entries. Values:\n",
       "  [1.0 0.0 … 0.0 0.0]\n",
       "  [0.0 1.0 … 0.0 0.0]\n",
       "  [0.0 0.0 … 0.0 0.0]\n",
       "  [0.0 0.0 … 0.0 0.0]\n",
       "  [0.0 0.0 … 1.0 0.0]\n",
       "  [0.0 0.0 … 0.0 1.0]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "values(vec_dict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1×6 RowVector{Float64,Array{Float64,1}}:\n",
       " 1.0  0.0  0.0  0.0  0.0  0.0"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vec_dict[\"programmer\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "6-element Array{Any,1}:\n",
       " [1.0 0.0 … 0.0 0.0]\n",
       " [0.0 1.0 … 0.0 0.0]\n",
       " [0.0 0.0 … 0.0 0.0]\n",
       " [0.0 0.0 … 0.0 0.0]\n",
       " [0.0 0.0 … 1.0 0.0]\n",
       " [0.0 0.0 … 0.0 1.0]"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "W = one_hot_vectorization(sample_text) # これが重みWになる"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.0"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "h = vec_dict[\"dog\"]*vec_dict[\"dog\"]'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Julia 0.6.2",
   "language": "julia",
   "name": "julia-0.6"
  },
  "language_info": {
   "file_extension": ".jl",
   "mimetype": "application/julia",
   "name": "julia",
   "version": "0.6.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}