# A novel hybrid model for emotion detection in text through sequential and transformer-based approaches: LSTM enhanced RoBERTa (LER)

## TL;DR

This paper introduces a new hybrid model for detecting emotions in text, combining LSTM and RoBERTa to achieve better accuracy than existing models.

## Contribution

The novel LSTM Enhanced RoBERTa (LER) model improves emotion detection by integrating sequential and transformer-based approaches.

## Key findings

- The LER model achieves 88% accuracy on the ISEAR emotion dataset.
- The hybrid model outperforms robust baseline models in emotion detection tasks.

## Abstract

Text emotion detection is an essential task in Natural Language Processing (NLP), with applications in customer support automation, diagnosing mental health, and social media analysis. Yet, precise emotion detection is a difficult problem as human emotional states are subtle, ambiguous, and contextual. Current models typically fail to fully grasp these subtleties. To overcome these limitations, the current study proposes a new hybrid architecture, LSTM Enhanced RoBERTa (LER), that combines the sequential learning ability of Long Short-Term Memory (LSTM) networks with the deep contextual knowledge provided by the RoBERTa transformer model. The LER model proposed here is tested on the popular ISEAR emotion dataset and registers an impressive accuracy of 88%, which surpasses many robust baseline models. The model’s performance is evaluated on standard metrics, such as precision, recall, and F1-score. The results show that the hybrid approach efficiently detects intricate emotional cues, thus improving the state of emotion detection for real-world, context-sensitive applications.

The online version contains supplementary material available at 10.1038/s41598-025-31984-1.