Prototipo de interfaz interactiva de análisis de emociones en textos

Abstract

For human beings, understanding emotions is not an easy task and it is often very complex to describe, interpret and evaluate the emotions we are feeling. Recognizing emotions is really important because this process is related to decision-making, which continually leads us to evaluate situations and identify possible sources of conflict and have control when initiating the response to those actions in order to resolve the conflict and generate an adequate response to the stimulus. On the other hand, natural language processing seeks to provide computers with the tools to understand, interpret and manipulate human language based on pattern recognition. In this project, a prototype of an interactive interface for emotion analysis in texts is developed, also a methodology is proposed and applied that allows the use of severarl text databases for the detection of basic emotions using NLP techniques and their integration with a graphical user interface that graphically indicates valence and emotional activation parameters. In addition, an artificial intelligence model is generated from the fine tuning of the BERT architecture for the classification of 4 emotions (joy, calm, anger and sadness) given by Russell's circumplex model of emotions with a weighted F1 score of 79%. Finally, the artificial intelligence model is integrated into a web interface on a local server for its use.