RequiBERT: Transfer Learning for Software Requirements Classifications

doi:10.1016/j.eswa.2026.132726

RequiBERT: Transfer Learning for Software Requirements Classifications

2026-05-03

0

PRE

AI

DOI：10.1016/j.eswa.2026.132726

Original

Original request for help

Save

Abstract

En 中文

• A pretrained domain-specific model, RequiBERT, is proposed to classify software requirements. • RequiBERT addresses the challenges of limited supervised datasets in requirements engineering. • To present the fine-tuning capability of RequiBERT with the small supervised dataset to create a state-of-the-art and robust model for software requirements classifications. • To analyze the robustness of RequiBERT on different classification tasks, including binary, a reduced set of NFR, and all NFR sets in the original PROMISE repository dataset.

Keywords:

RequiBERT

software requirements

transfer learning

classification

fine-tuning

RequiBERT: Transfer Learning for Software Requirements Classifications

Abstract

Journal

Organization