BIO-SUMM – UMA ESTRATÉGIA DE REDUÇÃO DE COMPLEXIDADE DE INFORMAÇÃO NÃO ESTRUTURADA

Abstract

The growing amounts of available online text and multimedia information in recent years have motivated the intensive research on automatic summarization methods. Automatic text summarization is the process where computer systems simulate the main features of humans to produce a condensed version of text documents with little or even no human intervention in the process. The abridged information that is generated (i.e., the summary) must also preserve the most relevant content of the text source according to the utilization purposes of the summary.
The use of summaries is as diverse as it is important; they are present in the routines of web users, in online information sources, in applications for mobile devices, as well as in corporate knowledge management systems. The approach proposed in this work consists of the formalization of a new methodology, called the Bio-Summ, which is based on the use of a modified version of the bio inspired Particle Swarm Clustering (PSC) algorithm in combination with other strategies to perform automatic text summarization. The effectiveness of the Bio-Summ is evaluated in a case study involving the summarization of 100 texts written in Brazilian Portuguese. The performance measures of the Bio-Summ are calculated automatically by a widely used summarization evaluation software system and are compared to five other summarization approaches.