gensim Latent Dirichlet Allocation minimum_probability vs print_topics

0
votes

My question is related to this post, Document topical distribution in Gensim LDA, the documentation for gensim.models.ldamodel states that "minimum_probability controls filtering the topics returned for a document (bow)." however, ldamodel[corpus] returns all possible topics with their probability (even below the number set in minimum_probability). what is the difference between these two? python 2.7.5 gensim 0.13.3

thank you

1
pythonldagensim
similar issue is raised in stackoverflow.com/a/36967532/5031727 but unlike the comment ldamodel[corpus] produces tuple of id and topics that are below the value set in min probability. – valearner

1 Answers

0
votes

My understanding of the documentation is that minimum_probability can be both the parameter of the model generation (applied to all queries afterwards) and the querying interface, e.g. get_document_topics(bow, minimum_probability=None, ....). So unless you trained with the minimum_probability parameter model[doc_bow] would not trim on probability.