This document describes LSI text clustering. It discusses vector space models, term weighting using TF-IDF, similarity measures, latent semantic indexing using singular value decomposition, suffix arrays and longest common prefix arrays for phrase discovery. The clustering algorithm involves preprocessing text, feature extraction to find terms and phrases, applying LSI to discover concepts and determine cluster labels, assigning documents to clusters, and calculating cluster scores. Parameters and issues with the algorithm are also outlined. A demo clusters a set of question and answer documents.