Description and analysis of contradiction for automatic multi-document summarization

• Grant number: 13/12521-5
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: September 01, 2013
• End date: August 31, 2014
• Field of knowledge: Linguistics, Literature and Arts - Linguistics - Linguistic Theory and Analysis
• Principal Investigator: Ariani Di Felippo
• Grantee: Naira Lícia da Silva
• Host Institution: Centro de Educação e Ciências Humanas (CECH). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil

Abstract

The Automatic Multi-document Summarization (MDS) is a sub-area of Natural Language Processing which aims at automatically producing a unique summary from a collection of texts, from different sources, on the same topic. The most widespread multi-document summaries are generic (i.e., aimed at a broad readership community) and informative (i.e., written to provide the main ideia of the original texts). In order to produce generic and informative multi-document summaries, MDS has very important challenges such as capturing the most important information of a topic within a generic perspective, eliminating redundancy and contradictory information. In particular, the identification of contradictory information is important because the absence of contradiction is one of the factors that ensure the informativeness of the summaries. The redundancy identification in MDS has been highlighted in several previous works. The contradiction, however, is a multi-document phenomenon that still needs to be addressed, at least in the MDS scenario. Thus, the goal of this project is to investigate the contradiction in the CSTNews corpus, which consists on a group of journalistic texts written in Portuguese language. Being part of the SUSTENTO project (2012/13246-5 FAPESP / CNPq 483231/2012-6), which aims at generating linguistic knowledge for MDS of Portuguese, the goal of this undergraduate research project is to describe and analyze contradiction in order to identify strategies that can support the detection of contradiction by automatic multi-document summarizers.