Numerical Claim Detection in Finance: A New Financial Dataset, Weak-Supervision Model, and Market Analysis

---

Overview | Citation | Contact

---

Overview

This repository contains the code for the paper "Numerical Claim Detection in Finance: A New Financial Dataset, Weak-Supervision Model, and Market Analysis".

The dataset and fine-tuned models are available on HuggingFace pages listed below. More information is available in the paper.

• Earnings call annotated data: HuggingFace Data
• RoBERTa-base model fine-tuned on earnings call data: HuggingFace Model
• RoBERTa-large model fine-tuned on analysts' reports data: HuggingFace Model

Citation

If you find this repository useful, please cite our work.

@article{shah2024numerical,
  title={Numerical Claim Detection in Finance: A New Financial Dataset, Weak-Supervision Model, and Market Analysis},
  author={Shah, Agam and Hiray, Arnav and Shah, Pratvi and Banerjee, Arkaprabha and Singh, Anushka and Eidnani, Dheeraj and Chava, Sahasra and Chaudhury, Bhaskar and Chava, Sudheer},
  journal={arXiv preprint arXiv:2402.11728},
  year={2024}
}

Contact

Please raise issue on GitHub or contact Agam Shah (ashah482[at]gatech[dot]edu) for any issues and questions.
GitHub: @shahagam4