Abstract:

The educational debate in social networks and its derivatives, such as bullying, has surfaced at the top of the academic agenda due to its impact on public opinion and, as a consequence, in the development and transformation plans of academic and governmental organizations. The scientific community has been working for several years to respond to the demand for tools that allow studying the educational attitude of citizens in these networks, focusing on sentiment analysis methodologies. However, their work has been hampered by several significant challenges. Great progress has been made in natural language processing (NLP) thanks to deep learning methodologies, but a correct analysis of the results also requires reaching a deep understanding of the "mental categories" of users. In other words, we are able to automatically process opinions, identifying their valence (positive, negative or neutral), but:

How can we base the subsequent categorization of concepts using a scientific procedure, according to the "mental model" of the users? Our purpose in this study is to propose a methodology to respond to this need of the research community, applying card sorting methodology and hierarchical clustering analysis.

This type of technique, borrowed from UX research, has been adapted and expanded for its application to a categorization model specifically oriented to the educational debate, in accordance with the objectives of the study.

Keywords:

Social media, mental models, education, sentiment analysis, card sorting, hierarchical clustering analysis, educational debate.