Improving Text Summarization Using Latent Semantic Analysis

Text Summarization is a method of generating a shorter version of the given document using natural language processing that enables the users to quickly identify the major points of a document. Text summarization aims at getting the most symbolic content in a system in a compact form from given document while it retains the semantic information of text to a large extent. It is considered to be an effective way of attempting the information and solves the problem of presenting information in more condense form. There are different approaches to produce well defined form of summaries and one of the modern methods is Latent Semantic Analysis. Though the available information about any topic is large and incredible, so there is a need for rapid view of those articles to determine accordance of the article as per user’s wish.

In this paper, the successive way of summarizing the text document by involving the sequence of the techniques and its evaluation using rouge scores was engaged. The SVD plays an important role in separating important sentences from input document. Every sentence is enabled with rank based on its importance in original document. Sentence selection is done based on their ranks and the summary generated. The rouge will produce three distinct scores as, Recall, Precision and F-score. The F-score is considered for evaluating the correctness of summary. The observation of three distinct summaries by reducing input document by 1/2nd, 1/3rd, 1/4th rouge scores and f-score is found to provide the effective results in summarizing the text document.