Visualizing and exploring multi-document summaries

• Grant number: 11/23200-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: March 01, 2012
• End date: December 31, 2013
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Thiago Alexandre Salgueiro Pardo
• Grantee: Jader Bruno Pereira Lima
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Associated scholarship(s): 12/16220-7 - Visualization and exploration of multi-document summaries, BE.EP.IC

Abstract

Computational applications that are capable of appropriately dealing with the incredible amount of available information nowadays, mainly on the web, are becoming increasingly more important. Text summarization is one of these applications, which aims at producing a unique summary from a group of documents on the same topic. There are several challenges to building a system like this, such as the suitable treatment of redundant, complementary, and contradictory information, the standardization of writing styles and referring expressions, the ordering of events/facts in time, and the maintenance of different foci and perspectives about the same event/fact, as well as the coherence and cohesion preservation in the summary. The way the summary is exhibited and the available possibilities for the user to explore it are as important as the summarization process itself. Usually, a user handling a summary automatically produced from a large group of texts (coming from Google or Google News, for instance) desires to be able to appropriately visualize this summary and to explore simple and more advanced functionalities, as to alternate between viewing the summary and the texts that contributed to some part of it, to look for some related information to some specific information in the summary, to compress even more or to expand the summary according to the user interests and available time, to filter the summary content based on some keywords and terms that the user may be interested in, to paint and comment some summary passages, to automatically translate the summary to some other language of interest, and to make available for other users a link to the summary, among several others. In this research plan we propose to investigate some of these functionalities and, based on them, to design a visual interface for the user to interact with the summary textual content, as well as to develop and evaluate it. Such proposal builds on the current research in multi-document summarization, enriching the products of these researches, at the same time that offers the student the opportunity to strengthen his experience in the scientific method. (AU)