Machine learning per a l'anàlisi i classificació de sentiments

Abstract

[en] Continuously people debate and express their opinions in their day to day life. Internet has encouraged these opinions to be made public through social networks or review platforms regarding a brand, product or service. To analyze this intake of information, mining of opinions is gaining strength. A scraping system has been implemented to obtain a real data corpus and algorithms based on natural language processing (NLP) and Machine Learning are used to implement a sentiment analysis framework. By means of supervised learning, the reviews written at TripAdvisor regarding tourist accommodation are classified. A brief introduction to the world of neuronal networks is made and the Keras library is used in a regression problem to predict scores. Finally, the result of analyzing opinions applied to hotels in the city of Barcelona is presented on a web map, obtaining an indicator that represents the degree of positive or negative comments for each hotel.