Home // BRAININFO 2025, The Tenth International Conference on Neuroscience and Cognitive Brain Information // View article

Predicting Emotion States Using Markov Chains

Authors:
Clement Leung
Zhifei Xu

Keywords: Image Emotion Prediction; Large Language Model; ChatGPT4; zero-shot; Markov Chain; Emotion Stability Parameter.

Abstract:
In a wide range of tasks, especially those involving critical safety considerations, it is crucial that human participants maintain appropriate emotional conditions. As a result, accurate recognition of these emotional states has become a central research challenge, with mainstream methods frequently utilizing Pre-trained Language Models (PLMs) to incorporate emotional understanding. With the emergence of Large Language Models (LLMs) like ChatGPT, we have seen remarkable advancements in various natural language processing applications. However, the potential of ChatGPT’s zero-shot capabilities for image-based emotion recognition and analysis has not been thoroughly explored. In this study, we focus on classifying and predicting emotional states, specifically distinguishing between positive and negative emotions, and we examine ChatGPT4’s ability to interpret emotions directly from images. Our experiments show that ChatGPT4 can effectively predict changes in emotional states over time, surpassing expectations in identifying the progression of positive and negative emotions. Nonetheless, we identified shortcomings in its capacity to accurately recognize specific negative emotions, indicating room for further improvement.

Pages: 7 to 16

Copyright: Copyright (c) IARIA, 2025

Publication date: March 9, 2025

Published in: conference

ISSN: 2519-8653

ISBN: 978-1-68558-239-5

Location: Lisbon, Portugal

Dates: from March 9, 2025 to March 13, 2025