Transforming Sentiment Analysis with a New LLM Architecture

Sentiment analysis from text is a critical task in the field of natural language processing, with wide-ranging applications in artificial intelligence and human-computer interaction. Emotions are physiological responses triggered by various experiences, and their analysis without relying on facial expressions or vocal cues requires supervised techniques to ensure accurate detection. Despite these challenges, understanding human emotions remains essential, especially as they are often expressed subtly through informal or inappropriate language on social platforms like Facebook and Twitter. In this study, we propose a deep learning-based system for emotion recognition. The system was evaluated on two distinct datasets: Tweeter_en_db in English and Snappfood in Persian. Recurrent neural networks and Long Short-Term Memory (LSTM) models were employed to demonstrate the system's capability of achieving high accuracy in emotion classification. Results indicate that our approach achieved ۹۰.۷۰% accuracy using a CNN model and ۸۸.۴۷% with LSTM on the English dataset, while on the Persian dataset, accuracy was ۸۲.۹۰% with CNN and ۸۵.۰۸% with LSTM. Comparative analysis shows that our methods outperform previous approaches by approximately ۸% on the Tweeter_en_db dataset and around ۲% on the Snappfood dataset.

کلیدواژه ها:

LLM ، LSTM ، CNN ، Sentiment