AAU-NLP/Cal-BERT-SL1000

Cal-BERT-SL1000

Model Description

Cal-BERT-SL1000 is a BERT-based sequence labeling model fine-tuned on the HiFi-KPI dataset for extracting financial key performance indicators (KPIs) from SEC earnings filings (10-K & 10-Q). It specializes in identifying entities that are one level up the calculation taxonomy ($n=1$), such as revenueAbstract, earnings, and financial ratios, using token classification.

This model was introduced in the paper HiFi-KPI: A Dataset for Hierarchical KPI Extraction from Earnings Filings by Rasmus Aavang, Giovanni Rizzi, Rasmus Bøggild, Alexandre Iolov, Mike Zhang (@jjzha), and Johannes Bjerva.

Use Cases

• Extracting financial KPIs using iXBRL calculation taxonomy
• Financial document parsing with entity recognition

Performance

• Trained on 1,000 most frequent labels from the HiFi-KPI dataset with $n=1$ in the calculation taxonomy.

Dataset & Code

• Dataset: HiFi-KPI on Hugging Face
• Official Code: HiFi-KPI GitHub Repository

Citation

@article{aavang2025hifikpi,
  title={HiFi-KPI: A Dataset for Hierarchical KPI Extraction from Earnings Filings},
  author={Aavang, Rasmus and Rizzi, Giovanni and B{\o}ggild, Rasmus and Iolov, Alexandre and Zhang, Mike and Bjerva, Johannes},
  journal={arXiv preprint arXiv:2502.15411},
  year={2025}
}