1 Andrew Polar, November 2, 2011 at 3:39 p.m.:

All these lectures look like unproductive activity of university professors in order to confirm their ability to derive complicated theoretical conclusions and having no impact on the benefits of society. I tested LSA and PLSA and does not have any advantage over Naive Bayes and Hierarchical Agglomerative Clustering. Details can be found on semanticsearchart.com. As university professors they should know that research should be backed up by theoretical proof of its advantage like it was done for Huffman encoding in data compression and Arithmetic encoding. It is theoretically proven that Huffman is better than Fano-Shannon and Arithmetic is better than Huffman. Show theoretically that LSA or PLSA better than NB, HAC, cosine similarity or other. That means show theoretically that it return better result statistically on all possible queries.