Implementação de melhorias no framework ShareMining para mineração de texto do Twitter

Abstract

In social networks, many data are generated in the form of text containing opinions on various topics. What you have to do with this information has been very prominent in recent times. Text mining is an area of data mining that involves extracting patterns and knowledge from text documents. Companies like Twitter or Facebook are able to generate relevant information with its profiles’ posts. How to do this analysis is a challenge for experts who often implement their own algorithms or use a variety of tools to process this data. Therefore, specialists can often implement the same algorithms more than once, making the process very manual and repetitive, since there are common steps when executing the text mining process, as has been identified in some works. Common steps for text mining can be summarized in collection, pre-processing, analysis and visualization. This work proposes improvements for Sharemining tool that implements all these steps in an automated and integrated way. Thus, the users can use more focus in the evaluation of the results than in the implementation of the whole process, making use of the framework. This paper also validates the improvements applied using Twitter data on the Brazilian social security reform.