README.md

---
language:
- "th"
tags:
- "thai"
- "token-classification"
- "pos"
- "dependency-parsing"
base_model: KoichiYasuoka/roberta-base-thai-char-upos
datasets:
- "universal_dependencies"
license: "apache-2.0"
pipeline_tag: "token-classification"
widget:
- text: "หลายหัวดีกว่าหัวเดียว"
---

# roberta-base-thai-char-ud-goeswith

## Model Description

This is a RoBERTa model pre-trained on Thai Wikipedia texts for POS-tagging and dependency-parsing (using `goeswith` for subwords), derived from [roberta-base-thai-char-upos](https://huggingface.co/KoichiYasuoka/roberta-base-thai-char-upos).

## How to Use

```py
from transformers import pipeline
nlp=pipeline("universal-dependencies","KoichiYasuoka/roberta-base-thai-char-ud-goeswith",trust_remote_code=True,aggregation_strategy="simple")
print(nlp("หลายหัวดีกว่าหัวเดียว"))
```

## Reference

Koichi Yasuoka: [Sequence-Labeling RoBERTa Model for Dependency-Parsing in Classical Chinese and Its Application to Vietnamese and Thai](https://doi.org/10.1109/ICBIR57571.2023.10147628), ICBIR 2023: 8th International Conference on Business and Industrial Research (May 2023), pp.169-173.