Commit History
-
Add VLSP2016 dataset support and comprehensive evaluation updates
Vu Anh +1: ClaudeMajor Features: - Add dataset selection parameter to train.py (--dataset vlsp2016|uts2017) - Train VLSP2016 models: Logistic Regression (70.19%), SVC Linear (71.14%) - Create clean.py script to remove runs without exported models - Update technical report with dual-dataset evaluation Technical Improvements: - Generic load_data() function routing to specific dataset loaders - Support for both UTS2017_Bank (35 aspect-sentiment classes) and VLSP2016 (3 sentiment classes) - N-gram comparison: bigrams (1-2) vs trigrams (1-3) analysis - Export filename includes dataset name for better organization Performance Results: - VLSP2016: 71.14% (SVC), 70.19% (LR) - balanced 3-class sentiment - UTS2017_Bank: 71.72% (SVC), 68.18% (LR) - imbalanced 35-class aspect-sentiment - Consistent ~71% accuracy across different complexity levels - Training efficiency: LR faster, SVC more accurate Documentation Updates: - Enhanced technical report with cross-dataset performance analysis - Added comparative insi…08bbb4c
-
Update technical report and README with latest SVC model results
Vu Anh +1: Claude- Updated technical report abstract and results with SVC accuracy: 71.72% - Added SVC algorithm details in classification model section - Updated README performance metrics with latest SVC results - Updated model filename references to latest exported model - Training accuracy: 94.31%, Test accuracy: 71.72%, Training time: 7.71s 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>0b1c1cf -
Remove sections 8 (Ethical Considerations) and 9 (Availability) from technical report
Vu Anh +1: Claude- Removed section 8: Ethical Considerations - Removed section 9: Availability - Document structure preserved with automatic section renumbering - Reduced document from 18 to 17 pages 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>93a8668 -
Update class count to exact number in README.md and technical report
Vu Anh +1: Claude- Change all instances of "35+" to "35" for precise class information - Updated README.md: UTS2017_Bank Dataset section header - Updated technical report: Abstract, introduction, datasets section, tables, and conclusion - Provides accurate information about model capabilities (exactly 35 aspect-sentiment combinations) - Verified from labels.txt file showing 35 unique classes 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>b0fc171 -
Remove PDF from git tracking and update project metadata
Vu Anh +1: Claude- Remove pulse_core_1_technical_report.pdf from git tracking (should be generated locally) - Update DOI in README.md citation (10.57967/hf/6605) - Update paper/.gitignore to exclude markdown files - PDF files should be generated locally from LaTeX source, not tracked in git 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>91cc050 -
Update author information format in technical report
Vu Anh +1: Claude- Move email from footnote to under organization name - Change from \thanks{Email: ...} to direct \texttt{email} format - Improve author block formatting: Name \\ Organization \\ Email 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>5240b04 -
Remove __pycache__ from git tracking and recover sample_papers files
Vu Anh +1: Claude- Removed __pycache__ directory from git tracking - Added .gitignore to exclude __pycache__ in future - Recovered sample_papers/arXiv-2509.04664v1/ files that were missing - Updated paper/.gitignore to exclude __pycache__ 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>264dfe0 -
Update Pulse Core 1 to focus on Vietnamese banking aspect sentiment analysis
Vu Anh +1: Claude- Updated README.md to reflect banking aspect sentiment focus with 68.18% accuracy - Transformed train.py to use UTS2017_Bank aspect sentiment dataset with <aspect>#<sentiment> labels - Updated inference.py to work with exported sentiment models - Updated technical report to remove VNTC references and focus on banking domain - Moved changelog to appendix and simplified to show 2025-09-28 as initial release - Added exported sentiment model and training artifacts - Removed VNTC dataset support entirely from codebase - Added LFS tracking for binary files 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>31e3523