Análise comparativa de algoritmos de construção de grafos e técnicas de incorporação de palavras na análise de sentimentos

Abstract

Sentiment analysis has become a crucial tool for understanding public perception in various areas, such as marketing, politics, and social media. It allows the extraction of valuable insights from large volumes of text, like consumer reviews or opinions expressed on social networks. Understanding the sentiment behind the words can guide business strategies, political campaigns, and even enhance user interaction. A common approach in sentiment analysis involves the application of machine learning techniques, which can range from simple rule-based methods to complex natural language processing models. Recently, with the advancement of artificial intelligence, more sophisticated methods have emerged that leverage not just textual content, but also the structural relationships of the data. With that said, the goal of this work is to conduct a comparative analysis of semi-supervised classification algorithms in graphs. These algorithms are particularly useful when there is a limited amount of labeled data available, a common situation in sentiment analysis due to the cost and effort required for manual annotation of large datasets. The experimental analysis explores the quality of graphs generated from different graph construction algorithms in relation to different word embeddings representations.