This post originated from an RSS feed registered with Ruby Buzz by Eigen Class. Original Post: Related document discovery, without algebra Feed Title: Eigenclass Feed URL: http://feeds.feedburner.com/eigenclass Feed Description: Ruby stuff --- trying to stay away from triviality. Latest Ruby Buzz Posts Latest Ruby Buzz Posts by Eigen Class Latest Posts From Eigenclass

You have heard about latent semantic analysis (if you haven't, take a look at this nice article on a SVD recommendation system written in 50 lines of Ruby). And you told yourself "hey, this is cool", to file it in your head right away. Or maybe you tried to actually use it, but were scared off by the algebra 101 part, or got lazy when you realized you needed to compile LAPACK, GSL or some other numerical library*1.

But you can get pretty far even without dimensionality reduction. If the feature space (e.g. the terms/concepts associated to your documents) is small enough, and you make sure synonymy is not a problem, you can do without algebra. One such case is that of your blog postings and their tags.

LSI is about reducing the dimensionality of a sparse term-document matrix, mitigating synonimy (different terms referring to the same idea) and polysemy (a word having multiple meanings). A program would do it using singular value decomposition, but you're also performing dimensionality reduction each time you tag your articles, mapping the text to a small number of keywords.

This means that you can use your tags to compute the cosine similarity between your posts, and find related pages in a dozen lines of code. The code that creates the "see also" box in eigenclass.org's pages looks essentially like this: