An Automatic Document Summarization System

Abstract

Nowadays digital documents often get collected in large numbers, due to high availability of them on the Internet and other storage and retrieval systems, such as modern libraries. Challenge of this digital era is to extract essential information from a document or collection of documents without lot of effort put into, going through entirety of it.

Rating of documents as per relevance determined based on user specified keywords is widely available and it allows users to reach sources containing relevant information. However, users will then need to look into details and facts inside the documents rated as relevant, before a document is really found to be useful. Often, there are instances where a large number of documents get returned with equal relevance rating, in response to a query. Such situations tell us the extent of the problem of information overload that we have to face at times. Summarizing is the process of dealing with a large amount of information by presenting and extracting only the essential material. There are numerous studies carried-out on generating of summaries of digital documents. Summarization approaches are broadly categorized as statistical methods and intelligent systems approaches and these are discussed in detail as part of this study on automatic document summarization. Further, different techniques used for automatic summarization are in the focus of the study. Conventional statistical methods used for automatic summary generation have been observed to be very useful over a period of time. In this study it is attempted to observe the same with development, testing and evaluation of a summarization system. System has the capability of specifying multiple documents for summarization and also capabilities to edit generated summary before it is saved into a separate file if necessary. Evaluation in the end explains how well system performs in its function of summary generation within the framework stipulated in the analysis and design of it.