Uma Análise das Principais Técnicas de Agrupamento de Dados, aplicadas a Coletâneas Textuais recuperadas de Páginas Web

Abstract

The volume of information is growing exponentially and, therefore, we should search appropriate solutions to extract knowledge from this huge mass. A suitable methodology for the handle large data sets is the Knowledge Discovery in Databases (KDD). This methodology consists of several steps and has Data Mining as main motor. However, we shouldn't employ the data mining tasks directly in the base, because, quite often, the tasks that make up this step doesn't deal with textual objects natively. To these sets of techniques used to interpret and analyze textual data it was called Knowledge Discovery in Texts (or Text Mining). This field, still emerging, intended for gathering meaningful information from texts in natural language, extracting knowledge from text documents. In this context, this research presents an overview of estages of KDD and its relationship with the Data Mining step, ilustrating some of its key methods through articles and scientific researches that employed them.In sequence, are detailed the steps of Text Mining, presenting their treatment and term's reduction tasks. For the first stage, the colection of documents, this work suggests the development of focused web crawlers in recovering contents from HTML pages, storing them in an easily processable format. After the theorical presentation, was applied the methodology studied to create a specific web crawler to the O Globo (digital edition) newspaper business section followed by the execution of clustering techniques to intuit about the main issues addressed in a certain period and check the possibility of grouping posts by topics and author