Ferramenta para Text Mining em Textos completos

Abstract

We live in a world in constant change where more and more people have unlimited access to the internet, where they can post their ideas. This means that there more and more texts available on the Web. These texts can be simple facebook posts or papers about various areas. With a large amount of texts, new techniques are needed to read and classify them quickly and effectively. Text Mining (TM) comes up to solve these problems. With this powerful tool it is possible to interpret and classify a huge amount of texts, in order to obtain useful information. Current TM approaches do not take advantage of structure of the texts. A text is seen as a bag of words, a set of unrelated words. This causes the TM algorithms become computationally heavy and their quality of information obtained can be substantially improved. There are different approaches of TM and several steps of preprocessing that can be applied to full text classification, but which methods that can get better results? Which steps must be applied so the final result is more complete? These are some questions that we address in this thesis.