Keyword extraction for document clustering using submodular optimization

With the rapid growth of information services, enormous amount of text corpus cannot simply be read and understand. Therefore, text clustering and visualization present a direct way to observe the documents as well as understand the topic by corresponding keywords. However, even a short paragraph contains a variety of words, which makes the keyword or topic extraction difficult to achieve. Therefore, we propose an algorithm to extract keywords efficiently and effectively, which makes use of the latent semantic indexing and submodular optimization. The visual layout allows users to simultaneously visualize (1) the overview of the whole dataset, (2) the detailed information in the specific scope of the collection of documents, and (3) the relationships of documents with their keywords.