So I am trying to use gensim to generate an LSI model along with corpus_lsi following this tutorial.
I start with a corpus and a dictionary that I generated myself. The list of documents are too small (9 lines = 9 documents), which is the sample list provided in gensim tutorials
However, pythos just crashes when it reaches the line for generating LSI_model. You can see below my code along with the generated output
Code
#!/usr/bin/env python
import os
from gensim import corpora, models, similarities
import logging
#logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.INFO)
if __name__ == '__main__':
if (os.path.exists("tmp\dictionary.dict")):
dictionary = corpora.Dictionary.load('tmp\dictionary.dict')
corpus = corpora.MmCorpus('tmp\corpus.mm')
print("Used files generated Dataset Generator")
else:
print("Please run dataset generator")
print ("generating tf-idf model ...")
tfidf = models.TfidfModel(corpus) # Generate tfidf matrix (tf-idf model)
print ("generating corpus_tf-idf model ...")
corpus_tfidf = tfidf[corpus] #use the model to transform vectors
print ("generating LSI model ...")
lsi = models.LsiModel(corpus_tfidf, id2word=dictionary, num_topics=2) # initialize an LSI transformation
print ("generating corpus_lsi model ...")
corpus_lsi = lsi[corpus_tfidf] # create a double wrapper over the original corpus: bow->tfidf->fold-in-lsi
lsi.print_topics(2)
Output
Used files generated Dataset Generator
generating tf-idf model ...
generating corpus_tf-idf model ...
generating LSI model ...
After printing "generating LSI model" it crashes
Any suggestions ?
Other things I tried
It seems that the issue was the function used in the tutorial (maybe downgraded or something)
so I changed the line
lsi = models.LsiModel(corpus_tfidf, id2word=dictionary, num_topics=2) # initialize an LSI transformation
To
lsi = LsiModel(corpus_tfidf,num_topics=2)
And it actually worked fine