Investigation of automatic multi-document summarization methods based on conceptual hierarchies

• Grant number: 14/12817-4
• Support Opportunities: Scholarships in Brazil - Master
• Start date: September 01, 2014
• End date: January 31, 2016
• Field of knowledge: Linguistics, Literature and Arts - Linguistics - Linguistic Theory and Analysis
• Agreement: Coordination of Improvement of Higher Education Personnel (CAPES)
• Principal Investigator: Gladis Maria de Barcellos Almeida
• Grantee: Andressa Caroline Inácio Zacarias
• Host Institution: Centro de Educação e Ciências Humanas (CECH). Universidade Federal de São Carlos (UFSCAR). São Carlos , SP, Brazil

Abstract

Automatic Multi-document Summarization (MDS) is an increasingly important task since there is a need to summarize document collections on the same topic to help users quickly find the most important information overall. In order to produce a unique informative and generic summary from a collection of texts, MDS has to select the main information of the collection that generic users might find important to include in a summary. To select the main information, there are content selection methods based on shallow and deep linguistic knowledge. Examples of deep methods are those using semantic-discursive or lexical conceptual information. Since there is no method based on lexical-conceptual knowledge for MDS of texts in Brazilian Portuguese language, we propose to investigate MDS methods in which the lexical concepts of the source texts are represented in a conceptual hierarchy, and the content selection is based on hierarchy properties that indicate the most important concepts of the collection. Thus, the content selection through conceptual hierarchy properties might produce informative automatic summaries. (AU)