Informativeness and Topicality in Multi-document summarization: new challenges and methods

• Grant number: 12/03071-3
• Support Opportunities: Regular Research Grants
• Start date: May 01, 2012
• End date: April 30, 2014
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Thiago Alexandre Salgueiro Pardo
• Grantee: Thiago Alexandre Salgueiro Pardo
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil

Abstract

Given the incredible and growing amount of available information, mainly on-line, and the difficulties and lack of time to deal with all this content, text processing applications have become increasingly relevant. Being relatively new (with origins in the mid-1990), multi-document summarization is one of these applications. It aims at automatically producing a unique summary from a group of texts on the same topic. Researches in the area started only in the last years in Brazil and for Portuguese language as well. From the creation of inedited resources and tools and the development of simple and naïve methods and systems to more sophisticated approaches, state of the art results were produced, and, in some cases, some of them were better than the ones obtained in international research and for other languages. Based on the recent research in the area, this research proposal aims at moving forwards and investigating 3 main correlated research questions that may advance the state of the art, namely: (i) how to jointly and appropriately deal with topicality in texts and summary informativeness; (ii) how to model and what is the impact of combining shallow/statistical and deep/linguistic methods in the production of more informative summaries that better mirror the topical distribution in the texts; and (iii) what are the characteristics of the summarization human/manual process that may be systematized and formalized in order to subsidize the previous research questions. While the first two questions deal with the production of better summaries, the last one may support the creation of new methods and propose different directions to the current approaches. Besides the training and qualification of human resources and the creation of a critical mass of researchers in the area, which is very small in Brazil, this project has potential to achieve significant contributions in the area by proposing innovative methods. (AU)