With the overwhelming amount of textual information available in electronic formats on the web, there is a need for an efficient text summarizer capable of condensing large bodies of text into shorter versions while keeping the relevant information intact. Such a technology would allow users to get their information in a shortened form, saving valuable time. Since 1997, Microsoft Word has included a summarizer for documents, and currently there are companies that summarize breaking news and send SMS for mobile phones. I wish to create a text summarizer to provide condensed versions of original documents. My focus is on blogs, because people are increasingly using this mode of communication to express their opinions on a variety of topics. Consequently, it will be very useful for a reader to be able to employ a concise summary, tailored to his or her own interests to quickly browse through volumes of opinions relevant to any number of topics. Although many summarization methods exist, my approach involves employing the Lanczos algorithm to compute eigenvalues and eigenvectors of a large sparse matrix and SVD (Singular Value Decomposition) as a means of identifying latent topics hidden in contexts; and the next phase of the process involves taking a high-dimensional set of data and reducing it to a lower-dimensional set. This procedure makes it possible to identify the best approximation of the original text. Since SQL makes it possible to allow analyzing data sets and take advantage of the parallel processing available today, in most database management systems, SQL is employed in my project. The utilization of SQL without external math libraries, however, adds to challenge in the computation of the SVD and the Lanczos algorithm.
Kim, Youn S., "Text Summarization" (2012). Master's Projects. 212.