Data descriptions
- Training set: 506 documents
- Development set: 250 documents
- Test set: 300 documents
Three named entities types: Location (LOC), Organiation (ORG), Person (PER)
Four relation types: LOCATED, PART-WHOLE, PERSONAL-SOCIAL, ORGANIZATION-AFFILIATION
| Model | F1 (micro-averaged) | F1 (macro-averaged) | Method | Reference | Code |
|---|---|---|---|---|---|
| RELEX2020_HT | 0.7206 | 0.6616 |
PhoBert,XML-RoBERTa underthesea |
Nguyen et al. VLSP'20 | |
| RELEX2020_MinhPQN | 0.6756 | 0.6342 |
NLPHust/vibert4news,FPT/vibert VnCoreNLP |
Pham VLSP'20 | |
| RELEX2020_SunBear | 0.6648 | 0.6209 |
join training NER and RE PhoBERT, VnCoreNLP |
Pham et al. VLSP'20 |
📜 Papers
- Nguyen et al. NICS'18. Relation Extraction in Vietnamese Text via Piecewise Convolution Neural Network with Word-Level Attention
- Truong et al. SoICT'17
- Sam et al. 2014
- Sam et al. SoICT'11
📁 Open sources
- vnoie (2017)
java