Comparative Study among Approaches based in Fuzzy Systems and Artificial Neural Networks to Estimate Importance of Comments about Products and Services

Abstract

The evolution of e-commerce and On-line Social Networks has contributed to the increase of the information available, making the task of analyzing the reviews manually almost impossible for the buying a product or service decisionmaking process. Due to the amount of information, the creation of automatic methods of knowledge extraction and data mining has become necessary. Currently, to facilitate the analysis of reviews some websites use filters such as votes by utility or by stars. However, the use of these filters is not a good practice because they may exclude reviews that have recently been submitted to the voting process, besides the possibility of the user overestimate or underestimate the review with attribution of stars. One possible solution is to filter the reviews based on their textual descriptions, author information and others measures. Sousa [1] proposed an approach, called TOP(X), to estimate the degree of importance of reviews using a Fuzzy System with three input variables: author reputation, extraction of tuples <feature, opinion word> and richness analyzer and an output variable: degree of importance of the review. Although the approach presented good results, some problems were pending of resolution and improvements, besides the possibility to change the computational model used. This work proposes adaptations in two input variables, namely: quantity of tuples and vocabulary richness and the building of new approaches using computational models based on Fuzzy Systems and Artificial Neural Networks (ANN). In addition, a comparison was made among the proposed approaches through statistical measures. Experiments performed in the hotel-domain showed that the approach using Fuzzy System obtained better results when detecting the most important reviews, without considering the semantic orientation of the comments. However, the approach using Multi-Layer Perceptron (MLP) Artificial Neural Networks obtained better results when is known the semantic orientation of the review.