Underthesea v9 (2026)

[rain1024]

Discussion thread about ideas and features of underthesea version 9 (2026)

Translation

• Vietnamese English and English Vietnamese Translation #854

Scikit-learn

• Incompatibility with sklearn >= 1.5 #731

Dependency Parse

• KeyError: 'reduce_ex_hook' #520
• ValueError: unknown url type: '/root/.underthesea/models/parsers/vi-dp-v1a0' #687
• KeyError: '__getitems__' #714
• Support PyTorch v2 for dependency parsing #706
• Update ParserTrainer #392
• Add new dependency parser with PhoBert #376

POS Tag

• Train model with VLSP2013-POS dataset #423

Dataset

• Create Underthesea Universal Dependency Dataset v0.1 #882

Dictionary

• Open Vietnamese Dictionary version 2023 #622

Rust

• Create CRF Library in Rust 🚀 #637
• Create Logistic Regression Library in Rust 🚀 #673

Python

• Support Python 3.14, deprecate Python 3.9 #861

Dependencies

• Remove NLTK Dependency #877

Done

• Visualize cây quan hệ với spacy #375
• Rust bindings in Docker on M1 #739
• The underthesea_core package on PyPI is missing the tar.gz package. #823
• AttributeError: jax.tree_map was removed in JAX v0.6.0: use jax.tree.map #762
• Install underthesea[wow] failed #676
• Lỗi gặp phải khi chuyển văn bản thành giọng nói (TTS) #727

Backlog

• Add Vietnamese diacritical marks #766
• Hỗ trợ phân tích tiếng Việt không dấu #200
• Migrate chatbot examples to Rasa 3.0 #526
• Create dictionary package for explore UVD  #525

Later

• Run underthesea in parallel programs #98
• Support Vietnamese TTS and ASR #173