BigFeel: um ambiente de processamento distribuído para integração de métodos de análise de sentimentos

Abstract

Sentiment analysis has been the main focus of plenty of research efforts, particularly justified by its commercial significance, both for consumers and businesses. Thus, many methods have been proposed, and the main ones have been compared in terms of effectiveness. Nonetheless, the literature is deficient when it comes to assessing the efficiency of these methods for processing large volumes of data, which are generated at great speed, volume and variety, known as Big Data. The present work presents an approach for integrating methods of sentiment analysis in order to process large volumes of data in a distributed environment, using both the Apache Hadoop and Spark platforms. A distributed application prototype was developed, named BigFeel, which supports the use of 22 methods of sentiment analysis, as well as some methods of natural language processing and textual preprocessing in large volumes of data. BigFeel offers services tailored to the use of computer networks, local and web, as well as offering an API for Scala/Java developers. The efficiency of the integrated methods was evaluated experimentally, demonstrating gain in comparison to the execution in the non-distributed implementation of the methods. Using the features offered by BigFeel, a case study of detection of innovation suggestions based on product and service reviews is also presented.