Replication Package for ReqSeek: Transformer-Based Automatic Requirements Identification for Building Large Requirements Datasets

📄 Usage Restriction Prior to Paper Publication
This repository is provided solely as a replication package for the peer review process. Do not use, modify, or redistribute any content until the paper is officially published. After publication, this repository will be archived on Zenodo with an open license.

---

Summary

This replication package supports our paper titled: ReqSeek: Transformer-Based Automatic Requirements Identification for Building Large Requirements Datasets, which introduces a transformer-based approach (ReqSeek) for automatically identifying requirements in textual data. Many deep-learning methods require large datasets, but collecting labeled requirements data is challenging. Our work aims to facilitate the creation of large-scale requirements datasets for training AI models.

---

Repository Content

1. ReqSeek/

   • This is the ReqSeek model; the best performing model. For more details, refer to the paper.

   • demo.ipynb: This notebook demonstrates how to download and use ReqSeek.

   • Note: To run ReqSeek, you might need GPU power

2. datasets/

   • Will contain our curated requirements datasets. The three curated datasets (ARID, SwaRD, UOCSSR) are exclusively submitted for review through the submission portal.
   • Includes a tutorial on how to load and use them.
   • We will make all three datasets publicly available, after the peer-review process is complete.

3. evaluations: Contains setup, experiments, and evaluation for all RQs.

---

🔒 License

Copyright © Mohammad Kasra Habib, 2025. All rights reserved.
Strictly no use, modification, or distribution permitted until paper publication.
Post-publication, this repository will be archived on Zenodo under an open license.