Utilizando um classificador de sentimento baseado em léxico em negociações de jogos de RPG para fazer predições no contexto de recursos naturais

Abstract

The management of natural resources focuses specifically on the scientific and technical understanding of these resources, their ecology, and the ability to support these resources’ lives. In this context, simulation appears as a mechanism that reproduces natural systems using computer tools to predict future consequences, to assist in decision making. Role-playing games are a type of game in which players interpret characters created within an environment, which organizes the players’ actions, determining the limits of what they can do or not. In the scope of water resources, RPG helps understand the problem and how each player thinks, analyzes, and observes the same from his perspective and thus elaborates the strategies he deems relevant to his role and the game. We developed this dissertation in the context of a project financed by CAPES and ANA, where an RPG game is under development for the participatory management of water resources. The analysis of feelings refers to natural language processing to systematically identify affective states and subjective information to create structured knowledge that a support system or decision-maker can use. We did not found studies that use techniques for analyzing feelings to add subjective attributes to predict behavior in games, in this case, to indicate pollution levels. The general objective of this dissertation is to construct models for the prediction of pollution levels with the use of subjective attributes through the integration of sentiment analysis techniques for the classification of transactions carried out by players. We created models using the theories of emotions proposed by Paul Ekman and Robert Plutchik. Through the evaluation metrics used, the built models show promising results. Throughout the RPG match, the models can predict the pollution levels considering the negotiation dialogues at the end of each round.