如何用gensim过滤掉语料库中低tf-idf的单词？

Question

I am using gensim for some NLP task. I've created a corpus from dictionary.doc2bow where dictionary is an object of corpora.Dictionary . Now I want to filter out the terms with low tf-idf values before running an LDA model. I looked into the documentation of the corpus class but cannot find a way to access the terms. Any ideas? Thank you.

Answer 1

Say your corpus is the following:

corpus = [dictionary.doc2bow(doc) for doc in documents]

After running TFIDF you can retrieve a list of low value words:

tfidf = TfidfModel(corpus, id2word=dictionary)

low_value = 0.2
low_value_words = []
for bow in corpus:
    low_value_words += [id for id, value in tfidf[bow] if value < low_value]

Then filter them out of the dictionary before running LDA:

dictionary.filter_tokens(bad_ids=low_value_words)

Recompute the corpus now that low value words are filtered out:

new_corpus = [dictionary.doc2bow(doc) for doc in documents]

Answer 2

This is old, but if you wanted to look at in on a per document level do something like this:

#same as before
dictionary = corpora.Dictionary(doc_list)
corpus = [dictionary.doc2bow(doc) for doc in doc_list]
tfidf = models.TfidfModel(corpus, id2word = dictionary)

#filter low value words
low_value = 0.025

for i in range(0, len(corpus)):
    bow = corpus[i]
    low_value_words = [] #reinitialize to be safe. You can skip this.
    low_value_words = [id for id, value in tfidf[bow] if value < low_value]
    new_bow = [b for b in bow if b[0] not in low_value_words]

    #reassign        
    corpus[i] = new_bow

Answer 3

This is essentially same as previous answers, but additionally handles words which are missing in tf-idf representation due to 0 score (terms present in all documents). Previous answer did not filter such terms and they still appeared in the final corpus.

#Same as before

dictionary = corpora.Dictionary(doc_list)
corpus = [dictionary.doc2bow(doc) for doc in doc_list]
tfidf = models.TfidfModel(corpus, id2word = dictionary)


#Filter low value words and also words missing in tfidf models.

low_value = 0.025

for i in range(0, len(corpus)):
    bow = corpus[i]
    low_value_words = [] #reinitialize to be safe. You can skip this.
    tfidf_ids = [id for id, value in tfidf[bow]]
    bow_ids = [id for id, value in bow]
    low_value_words = [id for id, value in tfidf[bow] if value < low_value]
    words_missing_in_tfidf = [id for id in bow_ids if id not in tfidf_ids] # The words with tf-idf socre 0 will be missing

    new_bow = [b for b in bow if b[0] not in low_value_words and b[0] not in words_missing_in_tfidf]  

#reassign        
corpus[i] = new_bow

Answer 4

Say you have a document tfidf_doc which generated by gensim's TfidfModel() with the corresponding bag of words document bow_doc , and you want to filter words that have tfidf value lower then cut_percent % of words in this document, you can call tfidf_filter(tfidf_doc, cut_percent) , then it will return a cut version of tfidf_doc :

def tfidf_filter(tfidf_doc, cut_percent):

    sorted_by_tfidf = sorted(tfidf_doc, key=lambda tup: tup[1])
    cut_value = sorted_by_tfidf[int(len(sorted_by_tfidf)*cut_percent)][1]

    #print('before cut:',len(tfidf_doc))

    #print('cut value:', cut_value)
    for i in range(len(tfidf_doc)-1, -1, -1):
        if tfidf_doc[i][1] < cut_value:
            tfidf_doc.pop(i)

    #print('after cut:',len(tfidf_doc))

    return tfidf_doc

Then you want to filter the document bow_doc by the resulting tfidf_doc , jsut call filter_bow_by_tfidf(bow_doc, tfidf_doc) , it will return cut version of bow_doc :

def filter_bow_by_tfidf(bow_doc, tfidf_doc):
    bow_idx = len(bow_doc)-1
    tfidf_idx = len(tfidf_doc)-1

    #print('before :', len(bow_doc))

    while True:
        if bow_idx < 0: break

        if tfidf_idx < 0:
            #print('pop2 :', bow_doc.pop(bow_idx))
            bow_doc.pop(bow_idx)
            bow_idx -= 1
        if bow_doc[bow_idx][0] > tfidf_doc[tfidf_idx][0]:
            #print('pop1 :', bow_doc.pop(bow_idx))
            bow_doc.pop(bow_idx)
            bow_idx -= 1
        if bow_doc[bow_idx][0] == tfidf_doc[tfidf_idx][0]:
            #print('keep :', bow_doc[bow_idx])
            bow_idx -= 1
            tfidf_idx -= 1

    #print('after :', len(bow_doc))

    return bow_doc