text.py

# Natural Language Toolkit: Texts
#
# Copyright (C) 2001-2021 NLTK Project
# Author: Steven Bird <stevenbird1@gmail.com>
#         Edward Loper <edloper@gmail.com>
# URL: <https://www.nltk.org/>
# For license information, see LICENSE.TXT

"""
This module brings together a variety of NLTK functionality for
text analysis, and provides simple, interactive interfaces.
Functionality includes: concordancing, collocation discovery,
regular expression search over tokenized strings, and
distributional similarity.
"""

import re
import sys
from collections import Counter, defaultdict, namedtuple
from functools import reduce
from math import log

from nltk.collocations import BigramCollocationFinder
from nltk.lm import MLE
from nltk.lm.preprocessing import padded_everygram_pipeline
from nltk.metrics import BigramAssocMeasures, f_measure
from nltk.probability import ConditionalFreqDist as CFD
from nltk.probability import FreqDist
from nltk.tokenize import sent_tokenize
from nltk.util import LazyConcatenation, tokenwrap

ConcordanceLine = namedtuple(
    "ConcordanceLine",
    ["left", "query", "right", "offset", "left_print", "right_print", "line"],
)


class ContextIndex:
    """
    A bidirectional index between words and their 'contexts' in a text.
    The context of a word is usually defined to be the words that occur
    in a fixed window around the word; but other definitions may also
    be used by providing a custom context function.
    """

    @staticmethod
    def _default_context(tokens, i):
        """One left token and one right token, normalized to lowercase"""
        left = tokens[i - 1].lower() if i != 0 else "*START*"
        right = tokens[i + 1].lower() if i != len(tokens) - 1 else "*END*"
        return (left, right)

    def __init__(self, tokens, context_func=None, filter=None, key=lambda x: x):
        self._key = key
        self._tokens = tokens
        if context_func:
            self._context_func = context_func
        else:
            self._context_func = self._default_context
        if filter:
            tokens = [t for t in tokens if filter(t)]
        self._word_to_contexts = CFD(
            (self._key(w), self._context_func(tokens, i)) for i, w in enumerate(tokens)
        )
        self._context_to_words = CFD(
            (self._context_func(tokens, i), self._key(w)) for i, w in enumerate(tokens)
        )

    def tokens(self):
        """
        :rtype: list(str)
        :return: The document that this context index was
            created from.
        """
        return self._tokens

    def word_similarity_dict(self, word):
        """
        Return a dictionary mapping from words to 'similarity scores,'
        indicating how often these two words occur in the same
        context.
        """
        word = self._key(word)
        word_contexts = set(self._word_to_contexts[word])

        scores = {}
        for w, w_contexts in self._word_to_contexts.items():
            scores[w] = f_measure(word_contexts, set(w_contexts))

        return scores

    def similar_words(self, word, n=20):
        scores = defaultdict(int)
        for c in self._word_to_contexts[self._key(word)]:
            for w in self._context_to_words[c]:
                if w != word:
                    scores[w] += (
                        self._context_to_words[c][word] * self._context_to_words[c][w]
                    )
        return sorted(scores, key=scores.get, reverse=True)[:n]

    def common_contexts(self, words, fail_on_unknown=False):
        """
        Find contexts where the specified words can all appear; and
        return a frequency distribution mapping each context to the
        number of times that context was used.

        :param words: The words used to seed the similarity search
        :type words: str
        :param fail_on_unknown: If true, then raise a value error if
            any of the given words do not occur at all in the index.
        """
        words = [self._key(w) for w in words]
        contexts = [set(self._word_to_contexts[w]) for w in words]
        empty = [words[i] for i in range(len(words)) if not contexts[i]]
        common = reduce(set.intersection, contexts)
        if empty and fail_on_unknown:
            raise ValueError("The following word(s) were not found:", " ".join(words))
        elif not common:
            # nothing in common -- just return an empty freqdist.
            return FreqDist()
        else:
            fd = FreqDist(
                c for w in words for c in self._word_to_contexts[w] if c in common
            )
            return fd


class ConcordanceIndex:
    """
    An index that can be used to look up the offset locations at which
    a given word occurs in a document.
    """

    def __init__(self, tokens, key=lambda x: x):
        """
        Construct a new concordance index.

        :param tokens: The document (list of tokens) that this
            concordance index was created from.  This list can be used
            to access the context of a given word occurrence.
        :param key: A function that maps each token to a normalized
            version that will be used as a key in the index.  E.g., if
            you use ``key=lambda s:s.lower()``, then the index will be
            case-insensitive.
        """
        self._tokens = tokens
        """The document (list of tokens) that this concordance index
           was created from."""

        self._key = key
        """Function mapping each token to an index key (or None)."""

        self._offsets = defaultdict(list)
        """Dictionary mapping words (or keys) to lists of offset indices."""
        # Initialize the index (self._offsets)
        for index, word in enumerate(tokens):
            word = self._key(word)
            self._offsets[word].append(index)

    def tokens(self):
        """
        :rtype: list(str)
        :return: The document that this concordance index was
            created from.
        """
        return self._tokens

    def offsets(self, word):
        """
        :rtype: list(int)
        :return: A list of the offset positions at which the given
            word occurs.  If a key function was specified for the
            index, then given word's key will be looked up.
        """
        word = self._key(word)
        return self._offsets[word]

    def __repr__(self):
        return "<ConcordanceIndex for %d tokens (%d types)>" % (
            len(self._tokens),
            len(self._offsets),
        )

    def find_concordance(self, word, width=80):
        """
        Find all concordance lines given the query word.

        Provided with a list of words, these will be found as a phrase.
        """
        if isinstance(word, list):
            phrase = word
        else:
            phrase = [word]

        half_width = (width - len(" ".join(phrase)) - 2) // 2
        context = width // 4  # approx number of words of context

        # Find the instances of the word to create the ConcordanceLine
        concordance_list = []
        offsets = self.offsets(phrase[0])
        for i, word in enumerate(phrase[1:]):
            word_offsets = {offset - i - 1 for offset in self.offsets(word)}
            offsets = sorted(word_offsets.intersection(offsets))
        if offsets:
            for i in offsets:
                query_word = " ".join(self._tokens[i : i + len(phrase)])
                # Find the context of query word.
                left_context = self._tokens[max(0, i - context) : i]
                right_context = self._tokens[i + len(phrase) : i + context]
                # Create the pretty lines with the query_word in the middle.
                left_print = " ".join(left_context)[-half_width:]
                right_print = " ".join(right_context)[:half_width]
                # The WYSIWYG line of the concordance.
                line_print = " ".join([left_print, query_word, right_print])
                # Create the ConcordanceLine
                concordance_line = ConcordanceLine(
                    left_context,
                    query_word,
                    right_context,
                    i,
                    left_print,
                    right_print,
                    line_print,
                )
                concordance_list.append(concordance_line)
        return concordance_list

    def print_concordance(self, word, width=80, lines=25):
        """
        Print concordance lines given the query word.
        :param word: The target word or phrase (a list of strings)
        :type word: str or list
        :param lines: The number of lines to display (default=25)
        :type lines: int
        :param width: The width of each line, in characters (default=80)
        :type width: int
        :param save: The option to save the concordance.
        :type save: bool
        """
        concordance_list = self.find_concordance(word, width=width)

        if not concordance_list:
            print("no matches")
        else:
            lines = min(lines, len(concordance_list))
            print(f"Displaying {lines} of {len(concordance_list)} matches:")
            for i, concordance_line in enumerate(concordance_list[:lines]):
                print(concordance_line.line)


class TokenSearcher:
    """
    A class that makes it easier to use regular expressions to search
    over tokenized strings.  The tokenized string is converted to a
    string where tokens are marked with angle brackets -- e.g.,
    ``'<the><window><is><still><open>'``.  The regular expression
    passed to the ``findall()`` method is modified to treat angle
    brackets as non-capturing parentheses, in addition to matching the
    token boundaries; and to have ``'.'`` not match the angle brackets.
    """

    def __init__(self, tokens):
        self._raw = "".join("<" + w + ">" for w in tokens)

    def findall(self, regexp):
        """
        Find instances of the regular expression in the text.
        The text is a list of tokens, and a regexp pattern to match
        a single token must be surrounded by angle brackets.  E.g.

        >>> from nltk.text import TokenSearcher
        >>> print('hack'); from nltk.book import text1, text5, text9
        hack...
        >>> text5.findall("<.*><.*><bro>")
        you rule bro; telling you bro; u twizted bro
        >>> text1.findall("<a>(<.*>)<man>")
        monied; nervous; dangerous; white; white; white; pious; queer; good;
        mature; white; Cape; great; wise; wise; butterless; white; fiendish;
        pale; furious; better; certain; complete; dismasted; younger; brave;
        brave; brave; brave
        >>> text9.findall("<th.*>{3,}")
        thread through those; the thought that; that the thing; the thing
        that; that that thing; through these than through; them that the;
        through the thick; them that they; thought that the

        :param regexp: A regular expression
        :type regexp: str
        """
        # preprocess the regular expression
        regexp = re.sub(r"\s", "", regexp)
        regexp = re.sub(r"<", "(?:<(?:", regexp)
        regexp = re.sub(r">", ")>)", regexp)
        regexp = re.sub(r"(?<!\\)\.", "[^>]", regexp)

        # perform the search
        hits = re.findall(regexp, self._raw)

        # Sanity check
        for h in hits:
            if not h.startswith("<") and h.endswith(">"):
                raise ValueError("Bad regexp for TokenSearcher.findall")

        # postprocess the output
        hits = [h[1:-1].split("><") for h in hits]
        return hits


class Text:
    """
    A wrapper around a sequence of simple (string) tokens, which is
    intended to support initial exploration of texts (via the
    interactive console).  Its methods perform a variety of analyses
    on the text's contexts (e.g., counting, concordancing, collocation
    discovery), and display the results.  If you wish to write a
    program which makes use of these analyses, then you should bypass
    the ``Text`` class, and use the appropriate analysis function or
    class directly instead.

    A ``Text`` is typically initialized from a given document or
    corpus.  E.g.:

    >>> import nltk.corpus
    >>> from nltk.text import Text
    >>> moby = Text(nltk.corpus.gutenberg.words('melville-moby_dick.txt'))

    """

    # This defeats lazy loading, but makes things faster.  This
    # *shouldn't* be necessary because the corpus view *should* be
    # doing intelligent caching, but without this it's running slow.
    # Look into whether the caching is working correctly.
    _COPY_TOKENS = True

    def __init__(self, tokens, name=None):
        """
        Create a Text object.

        :param tokens: The source text.
        :type tokens: sequence of str
        """
        if self._COPY_TOKENS:
            tokens = list(tokens)
        self.tokens = tokens

        if name:
            self.name = name
        elif "]" in tokens[:20]:
            end = tokens[:20].index("]")
            self.name = " ".join(str(tok) for tok in tokens[1:end])
        else:
            self.name = " ".join(str(tok) for tok in tokens[:8]) + "..."

    # ////////////////////////////////////////////////////////////
    # Support item & slice access
    # ////////////////////////////////////////////////////////////

    def __getitem__(self, i):
        return self.tokens[i]

    def __len__(self):
        return len(self.tokens)

    # ////////////////////////////////////////////////////////////
    # Interactive console methods
    # ////////////////////////////////////////////////////////////

    def concordance(self, word, width=79, lines=25):
        """
        Prints a concordance for ``word`` with the specified context window.
        Word matching is not case-sensitive.

        :param word: The target word or phrase (a list of strings)
        :type word: str or list
        :param width: The width of each line, in characters (default=80)
        :type width: int
        :param lines: The number of lines to display (default=25)
        :type lines: int

        :seealso: ``ConcordanceIndex``
        """
        if "_concordance_index" not in self.__dict__:
            self._concordance_index = ConcordanceIndex(
                self.tokens, key=lambda s: s.lower()
            )

        return self._concordance_index.print_concordance(word, width, lines)

    def concordance_list(self, word, width=79, lines=25):
        """
        Generate a concordance for ``word`` with the specified context window.
        Word matching is not case-sensitive.

        :param word: The target word or phrase (a list of strings)
        :type word: str or list
        :param width: The width of each line, in characters (default=80)
        :type width: int
        :param lines: The number of lines to display (default=25)
        :type lines: int

        :seealso: ``ConcordanceIndex``
        """
        if "_concordance_index" not in self.__dict__:
            self._concordance_index = ConcordanceIndex(
                self.tokens, key=lambda s: s.lower()
            )
        return self._concordance_index.find_concordance(word, width)[:lines]

    def collocation_list(self, num=20, window_size=2):
        """
        Return collocations derived from the text, ignoring stopwords.

            >>> from nltk.book import text4
            >>> text4.collocation_list()[:2]
            [('United', 'States'), ('fellow', 'citizens')]

        :param num: The maximum number of collocations to return.
        :type num: int
        :param window_size: The number of tokens spanned by a collocation (default=2)
        :type window_size: int
        :rtype: list(tuple(str, str))
        """
        if not (
            "_collocations" in self.__dict__
            and self._num == num
            and self._window_size == window_size
        ):
            self._num = num
            self._window_size = window_size

            # print("Building collocations list")
            from nltk.corpus import stopwords

            ignored_words = stopwords.words("english")
            finder = BigramCollocationFinder.from_words(self.tokens, window_size)
            finder.apply_freq_filter(2)
            finder.apply_word_filter(lambda w: len(w) < 3 or w.lower() in ignored_words)
            bigram_measures = BigramAssocMeasures()
            self._collocations = list(
                finder.nbest(bigram_measures.likelihood_ratio, num)
            )
        return self._collocations

    def collocations(self, num=20, window_size=2):
        """
        Print collocations derived from the text, ignoring stopwords.

            >>> from nltk.book import text4
            >>> text4.collocations() # doctest: +ELLIPSIS
            United States; fellow citizens; four years; ...

        :param num: The maximum number of collocations to print.
        :type num: int
        :param window_size: The number of tokens spanned by a collocation (default=2)
        :type window_size: int
        """

        collocation_strings = [
            w1 + " " + w2 for w1, w2 in self.collocation_list(num, window_size)
        ]
        print(tokenwrap(collocation_strings, separator="; "))

    def count(self, word):
        """
        Count the number of times this word appears in the text.
        """
        return self.tokens.count(word)

    def index(self, word):
        """
        Find the index of the first occurrence of the word in the text.
        """
        return self.tokens.index(word)

    def readability(self, method):
        # code from nltk_contrib.readability
        raise NotImplementedError

    def similar(self, word, num=20):
        """
        Distributional similarity: find other words which appear in the
        same contexts as the specified word; list most similar words first.

        :param word: The word used to seed the similarity search
        :type word: str
        :param num: The number of words to generate (default=20)
        :type num: int
        :seealso: ContextIndex.similar_words()
        """
        if "_word_context_index" not in self.__dict__:
            # print('Building word-context index...')
            self._word_context_index = ContextIndex(
                self.tokens, filter=lambda x: x.isalpha(), key=lambda s: s.lower()
            )

        # words = self._word_context_index.similar_words(word, num)

        word = word.lower()
        wci = self._word_context_index._word_to_contexts
        if word in wci.conditions():
            contexts = set(wci[word])
            fd = Counter(
                w
                for w in wci.conditions()
                for c in wci[w]
                if c in contexts and not w == word
            )
            words = [w for w, _ in fd.most_common(num)]
            print(tokenwrap(words))
        else:
            print("No matches")

    def common_contexts(self, words, num=20):
        """
        Find contexts where the specified words appear; list
        most frequent common contexts first.

        :param words: The words used to seed the similarity search
        :type words: str
        :param num: The number of words to generate (default=20)
        :type num: int
        :seealso: ContextIndex.common_contexts()
        """
        if "_word_context_index" not in self.__dict__:
            # print('Building word-context index...')
            self._word_context_index = ContextIndex(
                self.tokens, key=lambda s: s.lower()
            )

        try:
            fd = self._word_context_index.common_contexts(words, True)
            if not fd:
                print("No common contexts were found")
            else:
                ranked_contexts = [w for w, _ in fd.most_common(num)]
                print(tokenwrap(w1 + "_" + w2 for w1, w2 in ranked_contexts))

        except ValueError as e:
            print(e)

    def dispersion_plot(self, words):
        """
        Produce a plot showing the distribution of the words through the text.
        Requires pylab to be installed.

        :param words: The words to be plotted
        :type words: list(str)
        :seealso: nltk.draw.dispersion_plot()
        """
        from nltk.draw import dispersion_plot

        dispersion_plot(self, words)

    def _train_default_ngram_lm(self, tokenized_sents, n=3):
        train_data, padded_sents = padded_everygram_pipeline(n, tokenized_sents)
        model = MLE(order=n)
        model.fit(train_data, padded_sents)
        return model

    def generate(self, length=100, text_seed=None, random_seed=42):
        """
        Print random text, generated using a trigram language model.
        See also `help(nltk.lm)`.

        :param length: The length of text to generate (default=100)
        :type length: int

        :param text_seed: Generation can be conditioned on preceding context.
        :type text_seed: list(str)

        :param random_seed: A random seed or an instance of `random.Random`. If provided,
            makes the random sampling part of generation reproducible. (default=42)
        :type random_seed: int
        """
        # Create the model when using it the first time.
        self._tokenized_sents = [
            sent.split(" ") for sent in sent_tokenize(" ".join(self.tokens))
        ]
        if not hasattr(self, "_trigram_model"):
            print("Building ngram index...", file=sys.stderr)
            self._trigram_model = self._train_default_ngram_lm(
                self._tokenized_sents, n=3
            )

        generated_tokens = []

        assert length > 0, "The `length` must be more than 0."
        while len(generated_tokens) < length:
            for idx, token in enumerate(
                self._trigram_model.generate(
                    length, text_seed=text_seed, random_seed=random_seed
                )
            ):
                if token == "<s>":
                    continue
                if token == "</s>":
                    break
                generated_tokens.append(token)
            random_seed += 1

        prefix = " ".join(text_seed) + " " if text_seed else ""
        output_str = prefix + tokenwrap(generated_tokens[:length])
        print(output_str)
        return output_str

    def plot(self, *args):
        """
        See documentation for FreqDist.plot()
        :seealso: nltk.prob.FreqDist.plot()
        """
        return self.vocab().plot(*args)

    def vocab(self):
        """
        :seealso: nltk.prob.FreqDist
        """
        if "_vocab" not in self.__dict__:
            # print("Building vocabulary index...")
            self._vocab = FreqDist(self)
        return self._vocab

    def findall(self, regexp):
        """
        Find instances of the regular expression in the text.
        The text is a list of tokens, and a regexp pattern to match
        a single token must be surrounded by angle brackets.  E.g.

        >>> print('hack'); from nltk.book import text1, text5, text9
        hack...
        >>> text5.findall("<.*><.*><bro>")
        you rule bro; telling you bro; u twizted bro
        >>> text1.findall("<a>(<.*>)<man>")
        monied; nervous; dangerous; white; white; white; pious; queer; good;
        mature; white; Cape; great; wise; wise; butterless; white; fiendish;
        pale; furious; better; certain; complete; dismasted; younger; brave;
        brave; brave; brave
        >>> text9.findall("<th.*>{3,}")
        thread through those; the thought that; that the thing; the thing
        that; that that thing; through these than through; them that the;
        through the thick; them that they; thought that the

        :param regexp: A regular expression
        :type regexp: str
        """

        if "_token_searcher" not in self.__dict__:
            self._token_searcher = TokenSearcher(self)

        hits = self._token_searcher.findall(regexp)
        hits = [" ".join(h) for h in hits]
        print(tokenwrap(hits, "; "))

    # ////////////////////////////////////////////////////////////
    # Helper Methods
    # ////////////////////////////////////////////////////////////

    _CONTEXT_RE = re.compile(r"\w+|[\.\!\?]")

    def _context(self, tokens, i):
        """
        One left & one right token, both case-normalized.  Skip over
        non-sentence-final punctuation.  Used by the ``ContextIndex``
        that is created for ``similar()`` and ``common_contexts()``.
        """
        # Left context
        j = i - 1
        while j >= 0 and not self._CONTEXT_RE.match(tokens[j]):
            j -= 1
        left = tokens[j] if j != 0 else "*START*"

        # Right context
        j = i + 1
        while j < len(tokens) and not self._CONTEXT_RE.match(tokens[j]):
            j += 1
        right = tokens[j] if j != len(tokens) else "*END*"

        return (left, right)

    # ////////////////////////////////////////////////////////////
    # String Display
    # ////////////////////////////////////////////////////////////

    def __str__(self):
        return "<Text: %s>" % self.name

    def __repr__(self):
        return "<Text: %s>" % self.name


# Prototype only; this approach will be slow to load
class TextCollection(Text):
    """A collection of texts, which can be loaded with list of texts, or
    with a corpus consisting of one or more texts, and which supports
    counting, concordancing, collocation discovery, etc.  Initialize a
    TextCollection as follows:

    >>> import nltk.corpus
    >>> from nltk.text import TextCollection
    >>> print('hack'); from nltk.book import text1, text2, text3
    hack...
    >>> gutenberg = TextCollection(nltk.corpus.gutenberg)
    >>> mytexts = TextCollection([text1, text2, text3])

    Iterating over a TextCollection produces all the tokens of all the
    texts in order.
    """

    def __init__(self, source):
        if hasattr(source, "words"):  # bridge to the text corpus reader
            source = [source.words(f) for f in source.fileids()]

        self._texts = source
        Text.__init__(self, LazyConcatenation(source))
        self._idf_cache = {}

    def tf(self, term, text):
        """The frequency of the term in text."""
        return text.count(term) / len(text)

    def idf(self, term):
        """The number of texts in the corpus divided by the
        number of texts that the term appears in.
        If a term does not appear in the corpus, 0.0 is returned."""
        # idf values are cached for performance.
        idf = self._idf_cache.get(term)
        if idf is None:
            matches = len([True for text in self._texts if term in text])
            if len(self._texts) == 0:
                raise ValueError("IDF undefined for empty document collection")
            idf = log(len(self._texts) / matches) if matches else 0.0
            self._idf_cache[term] = idf
        return idf

    def tf_idf(self, term, text):
        return self.tf(term, text) * self.idf(term)


def demo():
    from nltk.corpus import brown

    text = Text(brown.words(categories="news"))
    print(text)
    print()
    print("Concordance:")
    text.concordance("news")
    print()
    print("Distributionally similar words:")
    text.similar("news")
    print()
    print("Collocations:")
    text.collocations()
    print()
    # print("Automatically generated text:")
    # text.generate()
    # print()
    print("Dispersion plot:")
    text.dispersion_plot(["news", "report", "said", "announced"])
    print()
    print("Vocabulary plot:")
    text.plot(50)
    print()
    print("Indexing:")
    print("text[3]:", text[3])
    print("text[3:5]:", text[3:5])
    print("text.vocab()['news']:", text.vocab()["news"])


if __name__ == "__main__":
    demo()

__all__ = [
    "ContextIndex",
    "ConcordanceIndex",
    "TokenSearcher",
    "Text",
    "TextCollection",
]