EUVIMEDEUVIMED entdecken
Crossref

Emotion Recognition from Text Using RoBERTa: A Deep Learning Approach for Enhanced Emotion Recognition

Irfanullah Safi, Fida Mohammad Safi, Mohammad Shah Omid

Bakhtar Journal of Engineering and Technology · 30.12.2025 · 1 (2) · 13

DOI: 10.65460/bjet/vol1_iss2_a2 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Bakhtar Journal of Engineering and Technology, online erschienen am 30.12.2025. Verfasst von Irfanullah Safi und 2 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Emotion Recognition in Conversations (ERC) has emerged as a crucial area in the development of emotionally intelligent systems. Recently, graph-based network (GBN) models have attracted significant interest for their ability to comprehend conversational contexts. However, these models frequently struggle to effectively capture and utilize contextual information in conversations, which hinders their overall effectiveness. To address these challenges, this study proposes a novel computational framework named AMRoBERTa (Augmentation Model A Robustly Optimized BERT Pretraining Approach). The model focuses on identifying human emotions by employing a balanced version of the Multimodal Emotion Lines Dataset (MELD). The approach leverages augmentation techniques to create multiple enriched datasets from MELD, which are subsequently used to train a deep neural network (DNN) integrated with transformer-based encoders. For text representation, the model utilizes RoBERTa (A Robustly Optimized Bidirectional Encoder Representations from Transformers), which ensures contextually rich word embeddings. By addressing the issue of data imbalance through dataset balancing, the proposed model enhances its capability to generalize across various scenarios. Comprehensive experimental evaluations on the MELD dataset reveal the model's superior performance, achieving a remarkable weighted F1-score of 66.20% and an accuracy of 66.20%, thereby outperforming existing methods in the field.

Zitieren

APA

Safi, I., Safi, F. M., & Omid, M. S. (2025). Emotion Recognition from Text Using RoBERTa: A Deep Learning Approach for Enhanced Emotion Recognition. Bakhtar Journal of Engineering and Technology, 1(2), 13. https://doi.org/10.65460/bjet/vol1_iss2_a2

Vancouver

Safi I, Safi FM, Omid MS. Emotion Recognition from Text Using RoBERTa: A Deep Learning Approach for Enhanced Emotion Recognition. Bakhtar Journal of Engineering and Technology. 2025;1(2):13. doi:10.65460/bjet/vol1_iss2_a2

BibTeX
@article{Safi2025,
  author = {Safi, Irfanullah and Safi, Fida Mohammad and Omid, Mohammad Shah},
  title = {Emotion Recognition from Text Using RoBERTa: A Deep Learning Approach for Enhanced Emotion Recognition},
  journal = {Bakhtar Journal of Engineering and Technology},
  year = {2025},
  volume = {1},
  number = {2},
  pages = {13},
  doi = {10.65460/bjet/vol1_iss2_a2}
}
Kostenlos starten →