This thesis is the result of a project whose objective has been to develop and deploy an
aspect-based sentiment analyzer applied to restaurant reviews based on Natural Language Processing (NLP) techniques, most of them developed at Intelligent Systems Group.
To do so, a system that extracts, contextualizes and classifies aspects was developed.
Specifically, each aspect commented by the user in his restaurant review.
For testing and evaluating the system we have used a dataset composed by restaurant
reviews that allows implementing each module in our analyzer. First, the system realizes
the extraction of aspects from each review. With the next module, the context of each
aspect is found. Once we have done this, the objective of next modules is to classify in
two ways these aspects. First, regarding six possible topics and finally depending on its
polarity, that is, if the aspect is valuated positive or negatively
Besides, a prototype has been validated with the dataset and a visualization system has
been developed using Web Components and D3.js to show analyzer results.
As a result, this project allows us to apply aspect-based sentiment analysis tasks to
restaurant reviews, which can lead an interesting study of restaurants market and lately
applying to other topics.