Search Result Clustering with Non Negative Matrix Factorisation

Non negative matrix factorisation. Wow, what a mouthful! Simply put, it's a technique to find two matrices (a features and weights matrix) that when multiplied together will reasonably approximate the original matrix.

Conveniently, the features matrix can be used to group or cluster rows in the original matrix that relate to terms or documents.

An article on codeproject explains it all in graphic detail...