Multi-document summarization using concept chain graphs
MetadataShow full item record
Automatic Text Summarization, especially, Multi-document Summarization, has received a great attention recently, due to the advent of web and the ease of creation and dissemination of digital content. We introduce some basic concepts related to Automatic Text Summarization and review the existing methods for Multi-document Summarization and its evaluation. The use of graphs for text representation to support Natural Language Processing is a relatively new idea. Concept Chain Graphs (CCG) is new content representation formalism which represents text as a hierarchical graph and is conducive for various Information Retrieval and Text Mining applications. We propose a new method for Multi-document Summarization based on CCG that uses well known Social Network Analysis (SNA) techniques to produce extracts from multiple documents and prove its effectiveness by experimental evaluation and comparison with other systems. The summarizer developed is intended to be a part of an application used to assist information analysts detect Unapparent Information Revelation (UIR) in a large pool of related documents. We provide a background of the application and discuss how this summarizer integrates with the application and increases its usefulness.