Estratégia para o problema de classificação de postagens relacionadas ao uso com aprendizagem baseada em regras

Abstract

With great and continuous technological advance in recent years it has become possible to popularization of computers and consequently social systems added new users, generating a large amount of data with the possibility of exploiting them. There are several specific forms and techniques of natural language processing (PLN) and machine learning to find some useful information from these data. This work focuses on the area of ​​Human Computer Interaction (UUX) and machine learning. In works been studied that the users of a specific system express how they feel about the when using them, usually with the use of postings, the main form of interaction of the majority of social systems. Such posts are called usage-related posts. (PRUs). Other works show a methodology that explores these PRUs and a UUX-POSTS, which supports the extraction and classification of the methodology, but with low hit rate. This paper proposes the use of a rules-based strategy for classification of the posts of a social network in PRU and non- PRU and compare the results with other strategies.