Reproduce “The Importance of Automatic Syntactic Features in Vietnamese Named Entity Recognition” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Thai-Hoang Pham, Phuong Le-Hong
A Bi-LSTM neural network augmented with automatically generated syntactic features significantly advances Vietnamese Named Entity Recognition, achieving a new state-of-the-art F1-score of 92.05% on the VLSP 2016 dataset. This performance represents an 18% F1 increase over models using only word embeddings, demonstrating the critical language-specific importance of these features for Vietnamese.
Key insights
- Automatic syntactic features (POS, chunk, regex) are exceptionally important for Vietnamese NER, providing an 18% F1-score boost when combined with word embeddings in a Bi-LSTM model.
- The importance of these syntactic features is language-specific, as they did not improve performance on the English CoNLL 2003 NER dataset, highlighting linguistic differences.
- A "simpler" deep learning architecture (2-layer Bi-LSTM) effectively augmented with linguistic features can outperform more complex end-to-end models for Vietnamese NER.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.