Hi, I have observed some issues with the compare functionality. I am using the extracted summaries and db from a very new wikipedia data (few months old). The article comparison model and other machine learning models that I am using remain the same as provided by the toolkit. I have not generated my own models. Now everything works fine, except that for some less related articles, I get quite high score (Reebok, Facebook) etc. For all related terms I get proper score. Also it doesn't happen for all the unrelated terms. So my guess is that the problem is with the article comparison model. Also one important thing to note is that the similarity score for the same unrelated terms are proper on the wikiminer site demos. Can someone please provide me some insight.
Log in to post a comment.