Instructions to use undertheseanlp/sen-1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Scikit-learn
How to use undertheseanlp/sen-1 with Scikit-learn:
from huggingface_hub import hf_hub_download import joblib model = joblib.load( hf_hub_download("undertheseanlp/sen-1", "sklearn_model.joblib") ) # only load pickle files from sources you trust # read more about it here https://skops.readthedocs.io/en/stable/persistence.html - Notebooks
- Google Colab
- Kaggle
Commit History
-
Refactor training to Hydra config and use underthesea imports
- Replace Click CLI with Hydra config system for flexible training - Add config files for all training tasks (vntc, bank, sentiment_general, sentiment_bank) - Change imports from underthesea_core to underthesea throughout - Move preprocessing to Rust TextPreprocessor (built into model binary) - Delete extends/ directory (code now in underthesea_core v3.2.0) - Add outputs/ to .gitignore for Hydra run outputs Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>903cdb2
-
Add sentiment models v1.2.0 with Vietnamese preprocessing
- Train sentiment-general (VLSP2016+UTS2017): 92.11% UTS2017, 70.86% VLSP2016 - Train sentiment-bank (UTS2017): 70.65% accuracy - Add preprocessing: lowercase, teencode expansion, negation marking, repeated char normalization - Update TECHNICAL_REPORT.md to v1.2.0 with full experiment results - Track .bin files with LFS/Xet storage Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>b5fd35d -
Add references folder and research skills
- Add 12 research papers (PhoBERT, ViSoBERT, vELECTRA, SMTCE, RoBERTa, XLM-R, UIT-VSMEC, UIT-VSFC, etc.) with PDF, LaTeX source, and Markdown - Add research synthesis: literature review, benchmark comparison, SOTA summary, bibliography - Add references website (index.html) with Vellum-style leaderboard, paper database, benchmarks, citation network, and research methodology guide - Add paper management tools (paper_db.py, fetch_papers.py, paper_db.json) - Add Claude Code skills for paper-fetch, paper-research, paper-review, paper-write - Configure git-lfs for binary files (PDF, PNG, etc.) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>ef06968
-
Upload models/sen-bank-1.0.0-20260203.bin with huggingface_hub
Verified signature1e215f7 -
Upload models/sen-general-1.0.0-20260203.bin with huggingface_hub
Verified signature5b958b7 -
Upload models/sen-bank-1.0.0-20260203.bin with huggingface_hub
Verified signature13e8419 -
Upload models/sen-general-1.0.0-20260203.bin with huggingface_hub
Verified signatureb88da25 -
Upload models/sen-bank-1.0.0-20260203.bin with huggingface_hub
Verified signature7667bf2 -
Upload models/sen-general-1.0.0-20260203.bin with huggingface_hub
Verified signaturea240e13 -
Upload models/sen-bank-1.0.0-20260203.bin with huggingface_hub
Verified signatured2a74e4 -
Upload models/sen-general-1.0.0-20260203.bin with huggingface_hub
Verified signatureea5c194 -
Update underthesea dependency to >=9.2.9
09af15a -
Add .env to .gitignore
0b026f1 -
Refactor: move Rust extensions to underthesea_core upstream
- Remove local Rust extensions (now in underthesea_core 3.1.7) - Consolidate training scripts into unified CLI (src/train.py) - Add benchmark CLI (src/bench.py) with vntc, bank, synthetic commands - Remove src/sen module (TextClassifier now in underthesea_core) - Add CLAUDE.md for project documentation - Update dependencies to use underthesea_core>=3.1.7 Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>b059f86
-
Add Rust extensions for TF-IDF and Linear SVM
- Implement TF-IDF vectorizer in pure Rust with PyO3 bindings - Implement Linear SVM with LIBLINEAR-style Dual Coordinate Descent - Achieve 88.15% accuracy on VNTC (vs sklearn's 89.48%) - Update text_classifier.py to use new Rust backend Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>2ac432e -
Update Python version to 3.10
21bb3e1 -
Add research plan and paper review documentation
- Add RESEARCH_PLAN.md: Phase 2 roadmap (PhoBERT comparison, word segmentation, ablation studies) - Add TECHNICAL_REPORT_REVIEW.md: ACL-style review with identified gaps - Update TECHNICAL_REPORT.md: Full benchmark results (92.49% VNTC, 75.76% Bank) - Update README.md and pyproject.toml Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>82c639c -
Initial commit: Sen-1 Vietnamese Text Classification
- src/sen: Core library (SenTextClassifier, Sentence, Label) - src/scripts: Training scripts (VNTC, UTS2017_Bank) - tests: Unit tests - Accuracy: 92.49% VNTC, 75.76% UTS2017_Bank - 17x faster than underthesea (single inference) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>c8e9ec2