Text and opinion mining techniques in social media environments

Resum

Social networks have assumed a crucial role in our lives, becoming a daily means of communication and information. This emergence has led to substantial advancements and improvements in various aspects of our daily routines. Social networks witness a massive influx of data every day, and when processed effectively, this data can confer competitive advantages to businesses and aid in the mitigation of significant issues for the society such as the proliferation of misinformation. This thesis focuses on the design and development of solutions specifically tailored to handle unstructured data from social networks, with a primary emphasis on opinion mining. The research has yielded promising results, including the introduction of unsupervised opinion mining techniques for large-scale sentiment analysis. Additionally, novel metrics and algorithms have been proposed to effectively combat misinformation, leveraging user-generated content and experience. The outcomes of this research have made substantial contributions to the field of opinion mining in social networks. The findings showcase significant progress in the analysis of unstructured data, coupled with effective strategies to counter the dissemination of misinformation and the study of opinion holders (users). These solutions provide robust and efficient means to comprehend the opinions and sentiments expressed within social networks, thereby presenting profound implications for both businesses and society as a whole.