Explainability in transformer-based models: methods, tools, and applications in natural language processing and bioinformatics
Briefings in Bioinformatics · 02.11.2026 · 27 (6)
DOI: 10.1093/bib/bbag447 ↗Kurz gefasst
Übersichtsarbeit in der Zeitschrift Briefings in Bioinformatics, online erschienen am 12.11.2026. Verfasst von Naeimeh M Karimi und 1 weiteren Autor:innen. Beteiligte Einrichtung: Department of Computer Engineering, Yazd University, Daneshgah Blvd., Safaeieh, Yazd, 8915818411, Iran. Hauptthemen laut MEDLINE: Natural Language Processing, Computational Biology. Die Arbeit nennt 122 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
In recent years, Transformer-based models have become state of the art in natural language processing (NLP) and have achieved exceptional performance in a wide range of tasks. Their impact has extended to critical domains such as healthcare and bioinformatics. Transformer models are nonlinear and complex, making it challenging for researchers to understand different aspects of model performance and decision-making processes. Given the widespread use of these models in data processing, particularly in applications related to human health, the need for interpretable and transparent models has become essential. The goal of this paper is to provide a comprehensive review of explainability methods for transformer models, as well as the interpretability tools commonly used in NLP and bioinformatics. First, interpretable methods in different domains are broadly categorized. It then presents a focused discussion on strategies for explaining transformer-based models. By analyzing and comparing these methods, their strengths, applicability to transformers, and limitations are highlighted. As model complexity increases, understanding model behavior becomes more challenging. Explainability techniques can help address this gap and support tasks such as error diagnostics, model improvement, and other applications. Greater clarity in a model’s decision-making process can increase user confidence. Interpretable models are particularly important in critical applications such as drug design, where they can support deeper investigation and facilitate new discoveries. Finally, this paper outlines potential directions for future research.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Computer Engineering, Yazd University, Daneshgah Blvd., Safaeieh, Yazd, 8915818411, Iran.
ORCID-Profile: Mehdi Rezaeian
Zitieren
APA
Karimi, N. M., & Rezaeian, M. (2026). Explainability in transformer-based models: methods, tools, and applications in natural language processing and bioinformatics. Briefings in Bioinformatics, 27(6). https://doi.org/10.1093/bib/bbag447
Vancouver
Karimi NM, Rezaeian M. Explainability in transformer-based models: methods, tools, and applications in natural language processing and bioinformatics. Briefings in Bioinformatics. 2026;27(6). doi:10.1093/bib/bbag447
BibTeX
@article{Karimi2026,
author = {Karimi, Naeimeh M and Rezaeian, Mehdi},
title = {Explainability in transformer-based models: methods, tools, and applications in natural language processing and bioinformatics},
journal = {Briefings in Bioinformatics},
year = {2026},
volume = {27},
number = {6},
doi = {10.1093/bib/bbag447}
}
EUVIMED