Open Vietnamese Natural Language Processing Experiments

Playground provides open experiments and tutorials for Vietnamese Natural Language Processing

Comparison with State-of-the-Art

Underthesea models compare with best methods on a range of Vietnamese NLP tasks:

Task	Dataset	Underthesea	Current/Previous Best
Classification	VNTC	92.78	97.01 (Vu RIVF'07)
Dependency Parsing	VLSP2020-DP	72.96 (LAS)	76.27 (Doan VLSP'20)

*Update: December 2020

Tutorials

We provide a set of experiments to get you build NLP models using underthesea:

• Training a Vietnamese Word Tokenizer
• Training a Vietnamese Text Classifier
• Training a Vietnamese Dependency Parser

Stack

• underthesea v1.3.2a1
• pytorch v1.5.1