Virtual environments such as online stores (e.g. Amazon, Google Play and Booking) adopt a collaborative strategy of evaluation and reputation, where users classify products and services. User's opinion represents the satisfaction level of a rated item. The set of ratings of an item is a reference to its reputation/quality. Therefore, the automatic identification of a usersatisfaction related to an item, considering its textual evaluation, is a tool with singular economic potential. With deep learning researches evolution in sentiment analysis based in aspects, opportunities to apply several neural networks in this context arisen. However, the data representation models applied in these works focus only on Embeddings pre-trained networks as a way to perform feature extraction. In this way, this work aims to present a comparison between data representation techniques and deep networks approaches, to analyze which of them have better results in classifying categories of aspects. Thus, we can seethat TF-IDF with a Convolution Neural Network (CNN) had an F1 measure of 0.93%, being at least 0.02% higher than the others approaches applied in this work.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.