Java implementation of a Term-Document matrix generator used as a tool to compute the Latent Semantic Index of the data set.
Screenshot of the command prompt that gets displayed when the generator is first ran.
Example of the .csv file that is generated based on a results pulling unique words from various documents.
We can load these matrices into matlab, and compute the SVD on these matrices. For example, if we want the top 6 query results from 12 documents, we can compute the average of the query term vectors and dot them with the query vector of each document.
Below is a screenshot of test query and the commands that were ran in matlab to compute the top 6 queries. In this specific example, the very top query is document 5. This is because the cosine similarity of document 5 is the closest to 1. Dissimilar documents have cosine similarities closer to 0.
