Emotion recognition and flip reasoning in english and mixed-coded con-versations based on a valence, arousal and dominance approach

Using the NRC VAD Lexicon and computational models like Transformer and GRU, this study presents a novel method for emotion recognition and reasoning about emo- tional transitions in code-mixed talks. We use Emotion Flip Reasoning (EFR) and Emotion Recognition in Conversation (ERC) to systematically identify and classify emotional triggers. We validate the model’s accuracy in identifying emotional shift triggers and classifying emotions using the MELD and MaSaC datasets provided by SemEval. Including VAD analysis significantly improves accuracy, as indicated by an increase in the F1 score. These findings open up new avenues for the study of emotional dynamics in texts with mixed codes by highlighting the significance of including complex emotional elements in conversation analysis. Finally, submit the results and manuscript to SemEval 2024 Competition.

• Autores: García Bedoya, Santiago Andre
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2024
• Institución: Universidad Tecnológica de Bolívar
• Repositorio: Repositorio Institucional UTB
• Idioma: eng
• OAI Identifier: oai:repositorio.utb.edu.co:20.500.12585/13916
• Acceso en línea: https://hdl.handle.net/20.500.12585/13916; https://utb.alma.exlibrisgroup.com/discovery/delivery/57UTB_INST:57UTB_INST/1242716950005731
• Palabra clave: Machine learning; Engineering -- Statistics; Computational linguistics; Python; Emotions; Mathematical statistics -- Data processing
• Rights: openAccess
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/