Natural Language Processing Techniques for Behavior Analysis in Social Networks of Hispanic American University Communities

Abstract

The Covid-19 confinement has forced certain human groups to look for alternatives to socialize. University communities did not stay out of this context. The presence of student communities called “confessions” whose purpose is to anonymously express their problems, desires and interests stands out. The main objective of this research is to determine the topics that highlight the cultural aspects and interests of these communities. Confessions pages from 5 Spanish-speaking countries were analyzed. Experiments were carried out on Facebookand Instagram posts using word embeddings and KMeans to cluster the semantic vector space. Anew evaluation approach based on the state-of-the-art was proposed that allow us to select and analyze topic models through specific keywords. As a result, topics of general interest were identified for each community such as “Academic life”, “Relationships”, “Politics” and “Personal problems”. The results vary by country. The collected dataset is publicly available1 for any academic purposes.