Análise de sentimentos usando centralidade de palavras baseada em feromônios de colônia de formigas

Abstract

Social networks have revolutionized the way we communicate and express our opinions. Platforms such as Facebook, Twitter and Instagram allow people to share their ideas, feelings and experiences in real time, creating a constant flow of information. This transformation has not only made it easier for individuals from different parts of the world to connect, but it has also democratized the production and consumption of content, giving voice to a diversity of perspectives that were previously less accessible. Sentiment analysis and text mining techniques allow researchers and companies to identify trends, patterns and emotions expressed in posts. This can be used for a variety of purposes, from improving marketing strategies to understanding social and political issues. This project develops an innovative methodology for analyzing sentiment in texts, using word centrality based on ant colony pheromones (ACO). The methodology involves data mining, building complex networks and applying ACO to identify word centrality. Inspired by the behavior of ants, the model allows a deep and contextualized analysis of feelings, where the pheromones that are deposited indicate emotional connections. The results of the experiments show that combining ACO with other centrality measures, such as Katz, PageRank, EigenVector and Degree Centrality, significantly improves the accuracy of sentiment analysis. Tests carried out on various datasets showed that the use of pheromones not only increases accuracy, but also the F1-Score. These results indicate that the proposed methodology is effective in capturing the emotional nuances present in texts. The practical application of this model can benefit various areas, such as marketing, customer service and social trend analysis, providing valuable insights into user behavior and preferences.