Minería de opinión a partir de las reseñas de TripAdvisor para analizarla percepción de los servicios que brindan los hoteles de la ciudad de Loja

Abstract

Abstract: Theobjective of this research work is to analyze and classify through machine learning techniques the reviews written by users of hotels in the city of Loja that are sent through the TripAdvisor application, in order to know the level of positivism or negativism.The process begins with the extraction of information, the reviews are then stored in a file to perform data cleaning and processing, in which all those characters that do not contribute to the analysis are eliminated and opinion mining is executed by applying machine learning methods. The GPT-3 rating code is identified and applied to calculate review polarity and rating scale. Each review of the MariaBD database is sent to a web service, where GPT-3 is in charge of sending the information to its server and the value of the analysis (positive, negative or neutral) is returned. Talend Open Studio is used to integrate the data, where a dynamic process is carried out: Talend -> Api in Nodejs -> OpenAI Api, it is the automatic bridge, to obtain better results. In the result, a classification of the model with an accuracy of 70% was obtained, when using the polarity classification code. On the other hand, when using the rating scale assignment method from 1 to 5, using the GPT-3 code, the level of precisionis 72%,for which the trend is given by the positive factor.