utils_nlp.py

import os
import random
import nltk
import wn


# Global singletones
wordnet = None


def get_word_synonyms(word, pos, limit: int = None):
    assert pos in 'anvr', "Invalid POS tag"
    global wordnet

    # Load the wordnet
    if wordnet is None:
        wordnet = wn.Wordnet('oewn:2023')

    # Find the list of synonyms
    synonyms = {word}
    synsets = wordnet.synsets(word, pos=pos)
    for syn in synsets:
        for sense in syn.senses():
            for lemma in sense.synset().lemmas():
                synonyms.add(lemma)
    synonyms.remove(word)
    synonyms = list(synonyms)

    # Return a subset
    if limit is not None:
        random.shuffle(synonyms)
        synonyms = synonyms[:limit]

    return synonyms


def upenn_to_wn_tag(upenn_tag: str, quiet=True):
    """
    Converts a UPenn tag to one of the four accepted tags by WordNet (a, v, n, r).
    Returns None if the tag is irrelevant in the context of WordNet.
    Prints a warning if the input tag is unavailable in UPenn tagset and quiet mode is disabled.
    More info on UPenn tags:
    ```
    import nltk
    nltk.help.upenn_tagset()
    ```
    """
    punctuation = "$`'(),-.:"
    irrelevant_tags = list(punctuation) + [
        'CC', # coordinating conjunction (and, both, ...)
        'CD', # numeral, cardinal
        'DT', # determiner
        'EX', # existential there
        'FW', # foreign word
        'IN', # preposition or conjunction, subordinating
        'LS', # list item marker
        'MD', # modal auxiliary
        'PDT', # pre-determiner
        'POS', # genitive marker
        'PRP', # pronoun, personal (hers, herself, him, himself, ...)
        'PRP$', # pronoun, possessive (her, his, mine, ...)
        'RP', # particle (aboard, about, across, upon, ...)
        'SYM', # symbol
        'TO', # "to" as preposition or infinitive marker
        'UH', # interjection (Goodbye, Goody, Gosh, Wow, ...)
        'WDT', # WH-determiner (that, what, whatever, ...)
        'WP', # WH-pronoun (that, what, whatever, whatsoever, ...)
        'WP$', # WH-pronoun, possessive (whose)
        'WRB', # Wh-adverb (how, however, whence, whenever, ...)
    ]
    adj_tags = [
        'JJ', # adjective or numeral, ordinal
        'JJR', # adjective, comparative
        'JJS', # adjective, superlative
    ]
    noun_tags = [
        'NN', # noun, common, singular or mass
        'NNP', # noun, proper, singular
        'NNPS', # noun, proper, plural
        'NNS', # noun, common, plural
    ]
    adv_tags = [
        'RB', # adverb
        'RBR', # adverb, comparative
        'RBS', # adverb, superlative
    ]
    verb_tags = [
        'VB', # verb, base form
        'VBD', # verb, past tense
        'VBG', # verb, present participle or gerund
        'VBN', # verb, past participle
        'VBP', # verb, present tense, not 3rd person singular
        'VBZ', # verb, present tense, 3rd person singular
    ]
    if upenn_tag in adj_tags:
        return 'a'
    elif upenn_tag in noun_tags:
        return 'n'
    elif upenn_tag in adv_tags:
        return 'r'
    elif upenn_tag in verb_tags:
        return 'v'
    elif upenn_tag in irrelevant_tags:
        return None
    else:
        if not quiet:
            # TODO: Log instead of print
            print(f'Warning: The given tag ({upenn_tag}) is not a valid UPenn tag.')
        return None

if __name__ == '__main__':
    # Download WordNet database
    os.system('python -m wn download oewn:2023')

    # Download NLTK packages
    nltk.download('averaged_perceptron_tagger')
    nltk.download('tagsets')
    nltk.download('punkt')