def get_embedding(paragraph, question, representation_name, vocab=None):
paragraphs_vectors = []
questions_vectors = []
if representation_name == "tfidf":
if vocab:
print("vocab specified")
vectorizer = TfidfVectorizer(vocabulary=vocab, ngram_range=(1, 3))
else:
vectorizer = TfidfVectorizer(max_features=15000,
ngram_range=(1, 3))
paragraphs_vectors, vectorizer = getTfIdfReprentation(
paragraph, vectorizer)
questions_vectors = vectorizer.transform(question) #.todense()
elif representation_name == "glove":
vectorizer = CountVectorizer()
X = vectorizer.fit(paragraph).vocabulary_
glove_dict = get_gloves_dict("glove.42B.300d.txt")
key_set = set(X.keys()) & set(glove_dict.keys())
glove_dict_vocab_corpus = {key: glove_dict[key] for key in key_set}
paragraphs_vectors = get_plong_corpus(paragraph,
glove_dict_vocab_corpus)
questions_vectors = get_plong_corpus(paragraph,
glove_dict_vocab_corpus)
elif representation_name == "bert":
vectorizer = Vectorizer()
vectorizer.bert(paragraph)
paragraphs_vectors = vectorizer.vectors
vectorizer.bert(question)
questions_vectors = vectorizer.vectors
else:
print("error representation_name")
return paragraphs_vectors, questions_vectors
