Deep Learning Approach for Aspect-Based Sentiment Analysis of Restaurants Reviews in Spanish

Abstract

Online reviews of products and services have become important for customers and enterprises. Recent research focuses on analyzing and managing those kinds of reviews using natural language processing. This paper focuses on aspect-based sentiment analysis for reviews in Spanish. First, the reviews data sets are normalized into different inputs of the neural networks. Then, our approach combines two deep learning models architectures to determine a positive or negative assessment and identify the most important characteristics or aspects of the text. We develop two architectures for aspect detection and three architectures for sentiment analysis. Merging the deep learning models, we tested our approach in restaurant reviews and compared them with state-of-the-art methods.