File size: 1,271 Bytes
278590a
 
 
a5ea543
 
 
 
 
 
 
 
 
6c01d6c
1351cea
 
 
3f7fc4c
 
f1f2c64
8559d2a
bb8e9cb
 
f1a89aa
 
bb8e9cb
 
 
 
 
f1a89aa
bb8e9cb
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
---
base_model:
- jjzha/esco-xlm-roberta-large
license: apache-2.0
language:
- en
pipeline_tag: text-classification
tags:
- text-classification
- skill-detection
- sentence-classification
- ESCO
library_name: transformers
metrics:
- f1
- accuracy
---

We fine-tune jjzha/esco-xlm-roberta-large for sentence-level binary skill identification. The results show 94% accuracy and F1 score in English. Furthermore, the study demonstrates the model's effectiveness for cross-lingual transfer. Please refer to the original paper for more information, and if you use this work, please cite the following:

Musazade, N., Zhang, M., & Mezei, J. (2025, August). Cross-Lingual Sentence-Level Skill Identification in English and Danish Job Advertisements. In Proceedings of the 8th International Conference on Natural Language and Speech Processing (ICNLSP-2025) (pp. 410-415).

https://aclanthology.org/2025.icnlsp-1.40.pdf 

@inproceedings{musazade2025cross,
  title={Cross-Lingual Sentence-Level Skill Identification in English and Danish Job Advertisements},
  author={Musazade, Nurlan and Zhang, Mike and Mezei, J{\'o}zsef},
  booktitle={Proceedings of the 8th International Conference on Natural Language and Speech Processing (ICNLSP-2025)},
  pages={410--415},
  year={2025}
}