Coverage for nltk.chunk.api : 60%

# Natural Language Toolkit: Chunk parsing API 

# 

# Copyright (C) 2001-2012 NLTK Project 

# Author: Edward Loper <edloper@gradient.cis.upenn.edu> 

#         Steven Bird <sb@csse.unimelb.edu.au> (minor additions) 

# URL: <http://www.nltk.org/> 

# For license information, see LICENSE.TXT 

##////////////////////////////////////////////////////// 

##  Chunk Parser Interface 

##////////////////////////////////////////////////////// 

from nltk.parse import ParserI 

from nltk.chunk.util import ChunkScore 

class ChunkParserI(ParserI): 

    """ 

    A processing interface for identifying non-overlapping groups in 

    unrestricted text.  Typically, chunk parsers are used to find base 

    syntactic constituents, such as base noun phrases.  Unlike 

    ``ParserI``, ``ChunkParserI`` guarantees that the ``parse()`` method 

    will always generate a parse. 

    """ 

    def parse(self, tokens): 

        """ 

        Return the best chunk structure for the given tokens 

        and return a tree. 

        :param tokens: The list of (word, tag) tokens to be chunked. 

        :type tokens: list(tuple) 

        :rtype: Tree 

        """ 

        raise NotImplementedError() 

    def evaluate(self, gold): 

        """ 

        Score the accuracy of the chunker against the gold standard. 

        Remove the chunking the gold standard text, rechunk it using 

        the chunker, and return a ``ChunkScore`` object 

        reflecting the performance of this chunk peraser. 

        :type gold: list(Tree) 

        :param gold: The list of chunked sentences to score the chunker on. 

        :rtype: ChunkScore 

        """ 

        chunkscore = ChunkScore() 

        for correct in gold: 

            chunkscore.score(correct, self.parse(correct.leaves())) 

        return chunkscore