try:
from nltk.corpus import wordnet as wn
raise_lookuperror_if_wordnet_data_absent = wn.synsets("python")
except LookupError:
import nltk
nltk.download("wordnet")
import inflect
from .constants import (ALL_WORDNET_WORDS, CONJUGATED_VERB_LIST,
ADJECTIVE_TO_ADVERB)
def belongs(lemma, lemma_list):
"""
args:
- lemma : a Wordnet lemma e.g. Lemma('administration.n.02.governance')
- lemma_list : a list of lemmas
returns True if lemma is in lemma_list and False otherwise.
The NLTK Wordnet implementation of Lemma is such that two lemmas are
considered equal only if their names are the same.
i.e Lemma('regulate.v.02.govern') == Lemma('govern.v.02.govern')
Therefore the python statement "lemma in lemma_list" yields
unexpected results. This function implements the expected
behavior for the statement "lemma in list_list".
"""
for element in lemma_list:
if (element.name() == lemma.name() and
element.synset() == lemma.synset()):
return True
return False
def get_related_lemmas(word):
"""
args
- word : a word e.g. "lovely"
returns a list of related lemmas e.g [Lemma('cover_girl.n.01.lovely'),
Lemma('lovely.s.01.lovely'), Lemma('adorable.s.01.lovely'),
Lemma('comeliness.n.01.loveliness')]
returns [] if Wordnet doesn't recognize the word
"""
all_lemmas_for_this_word = [lemma for ss in wn.synsets(word)
for lemma in ss.lemmas()
if lemma.name() == word]
all_related_lemmas = [lemma for lemma in all_lemmas_for_this_word]
new_lemmas = []
for lemma in all_lemmas_for_this_word:
for new_lemma in (lemma.derivationally_related_forms() +
lemma.pertainyms()):
if (not belongs(new_lemma, all_related_lemmas) and
not belongs(new_lemma, new_lemmas)):
new_lemmas.append(new_lemma)
while len(new_lemmas) > 0:
all_lemmas_for_new_words = []
for new_lemma in new_lemmas:
word = new_lemma.name()
all_lemmas_for_this_word = [lemma for ss in wn.synsets(word)
for lemma in ss.lemmas()
if lemma.name() == word]
for lemma in all_lemmas_for_this_word:
if not belongs(lemma, all_lemmas_for_new_words):
all_lemmas_for_new_words.append(lemma)
all_related_lemmas += all_lemmas_for_new_words
new_lemmas = []
for lemma in all_lemmas_for_new_words:
for new_lemma in (lemma.derivationally_related_forms() +
lemma.pertainyms()):
if (not belongs(new_lemma, all_related_lemmas) and
not belongs(new_lemma, new_lemmas)):
new_lemmas.append(new_lemma)
return all_related_lemmas
def singularize(noun):
"""
args
- noun : a noun e.g "man"
returns the singular form of the word if it finds one. Otherwise,
returns the word itself.
"""
singular = inflect.engine().singular_noun(noun)
if singular in ALL_WORDNET_WORDS:
return singular
return noun
def get_word_forms(word):
"""
args
word : a word e.g "love"
returns the related word forms corresponding to the input word. the output
is a dictionary with four keys "n" (noun), "a" (adjective), "v" (verb)
and "r" (adverb). The value for each key is a python Set containing
related word forms with that part of speech.
e.g. {'a': {'lovable', 'loveable'},
'n': {'love', 'lover', 'lovers', 'loves'},
'r': set(),
'v': {'love', 'loved', 'loves', 'loving'}}
"""
word = singularize(word)
related_lemmas = get_related_lemmas(word)
related_words_dict = {"n" : set(), "a" : set(), "v" : set(), "r" : set()}
for lemma in related_lemmas:
pos = lemma.synset().pos()
if pos == "s":
pos = "a"
related_words_dict[pos].add(lemma.name())
noun_set = [noun for noun in related_words_dict["n"]]
for noun in noun_set:
related_words_dict["n"].add(inflect.engine().plural_noun(noun))
verb_set = [verb for verb in related_words_dict["v"]]
for verb in verb_set:
for conjugated_verbs in CONJUGATED_VERB_LIST:
if verb in conjugated_verbs:
for conjugated_verb in conjugated_verbs:
related_words_dict["v"].add(conjugated_verb)
adjective_set = [adjective for adjective in related_words_dict["a"]]
for adjective in adjective_set:
try:
related_words_dict["r"].add(ADJECTIVE_TO_ADVERB[adjective])
except KeyError:
pass
return related_words_dict
