README.md

---
license: cc-by-nc-sa-4.0
language: de
---

This is the skill dataset created by:

```
@inproceedings{gnehm-etal-2022-fine,
    title = "Fine-Grained Extraction and Classification of Skill Requirements in {G}erman-Speaking Job Ads",
    author = {Gnehm, Ann-sophie  and
      B{\"u}hlmann, Eva  and
      Buchs, Helen  and
      Clematide, Simon},
    booktitle = "Proceedings of the Fifth Workshop on Natural Language Processing and Computational Social Science (NLP+CSS)",
    month = nov,
    year = "2022",
    address = "Abu Dhabi, UAE",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2022.nlpcss-1.2",
    doi = "10.18653/v1/2022.nlpcss-1.2",
    pages = "14--24",
}
```

There are document delimiters indicated by `idx`.

Number of samples (sentences):
- train: 19889
- dev: 2332
- test: 2557

Sources:
- Swiss Job Market Monitor (SJMM): https://www.swissubase.ch/en/

Type of tags:
- BI(-ICT) and O tags with keys `tags_skill`

Sample:
```
{
    "idx": 198, 
    "tokens": ["-", "besitzen", "fundierte", "Anwenderkenntnisse", "in", "MS-Office"], 
    "tags_skill": ["O", "O", "O", "O", "O", "B-ICT"]
}
```