Re:Text Matching Using Cosine Similarity
by autokad on Thursday December 27, 2018 @11:03AM
(#18770238)
its nice to use when you don't have labeled data such as plagiarism detection. I use it a lot in the security space to see if a user has changed their behaviors by looking at the cosine distance of resources used over a period of time vs the next.
Right now on the front page there is another article about "deep learning" classifiers which are competitive with this technology and it would be nice to see an objective comparison. (e.g. a co-worker made a 0.93 accurate text classifier in 2004 w/ bag o' words and the SVM)