---
license: mit
task_categories:
- text-classification
language:
- en
tags:
- Skill Extraction
pretty_name: Skill Extraction - TechWolf
size_categories:
- n<1K
---
# Skill Extraction with ESCO skills - TechWolf subset

## Dataset Description

- **Paper:** https://arxiv.org/abs/2307.10778
- **Point of Contact:** jensjoris@techwolf.ai

## Dataset Summary

The `TECHWOLF` subset, although smaller, represents a more generic distribution of job descriptions and skill spans. [ESCO](https://esco.ec.europa.eu/en/classification/skill_main) skills are directly annotated on the full sentence level, thus omitting the intermediate span identification step. ESCO v1.1.0 is used.

This dataset is part of a three-part evaluation dataset for skill extraction:
1. [**skill-extraction-tech**](https://huggingface.co/datasets/jensjorisdecorte/skill-extraction-tech)
2. [**skill-extraction-house**](https://huggingface.co/datasets/jensjorisdecorte/skill-extraction-house)
3. [**skill-extraction-techwolf**](https://huggingface.co/datasets/jensjorisdecorte/skill-extraction-techwolf)

### Citation Information

If you use this dataset, please include the following reference:

```
@article{decorte2023extreme,
  title={Extreme multi-label skill extraction training using large language models},
  author={Decorte, Jens-Joris and Verlinden, Severine and Van Hautte, Jeroen and Deleu, Johannes and Develder, Chris and Demeester, Thomas},
  journal={arXiv preprint arXiv:2307.10778},
  year={2023}
}
```