Query-Classifier

A BERT-based classifier that distinguishes between Gmail and Calendar queries. For calendar queries, it can also extract a time range from the user's input using natural language expressions.

---

Features

• Classifies queries as either gmail or calendar.
• Extracts time ranges from calendar queries using natural language expressions.
• Interactive widget for live testing of queries.
• Automated evaluation with detailed test case reporting.
• Confusion matrix and metrics for model performance.

---

System Specs

• OS: Arch Linux

• CPU: AMD Ryzen 7 7840HS

• GPU: NVIDIA GeForce RTX 4060

• RAM: 32GB

• Python: 3.10.X

• CUDA: 12.9

• PyTorch: 2.5.1, (GPU-CUDA) 11.8

Python Libraries

• torch
• transformers
• pandas
• scikit-learn
• tqdm
• seaborn
• matplotlib
• dateparser
• ipywidgets

Install all dependencies with:

pip install torch transformers pandas scikit-learn tqdm seaborn matplotlib dateparser ipywidgets

---

Usage

1. Prepare Data:
   Place your labeled queries in labeled-queries.csv (see example in the repo).

2. Run the Notebook:
   Open queryclassifier-final.ipynb or upgrade.ipynb in Jupyter or VS Code.

3. Train the Model:
   The notebook will guide you through data loading, training, and evaluation.

4. Test the Model:

   • Use the interactive widget to test custom queries.
   • Run the automated test cell to evaluate on test-set.csv.

---

File Structure

• queryclassifier-final.ipynb — Main notebook for training, evaluation, and testing.
• labeled-queries.csv — Labeled data for training/validation/testing.
• test-set.csv — Diverse test cases for evaluation.
• README.md — Project documentation.

---

Example Queries

• Gmail:

  • "Show me my emails"
  • "Find emails from Alice"
  • "Did I get any mail today?"

• Calendar:

  • "What events do I have this week?"
  • "Show calendar for June 2025"
  • "What's on my agenda?"

---

Notes

• The time extraction utility is rule-based and works for common time expressions (e.g., "this week", "next month", "June 2025").
• Ambiguous or mixed queries (e.g., "Show me my calendar and emails") are challenging and may require more advanced modeling or labeling.
• For best results, use a GPU for training.

Note:

The interactive widget for live query testing requires the ipywidgets library and a Jupyter Notebook environment (such as JupyterLab, Jupyter Notebook, or VS Code with the Jupyter extension).

If you run the notebook in a plain terminal or an environment that does not support widgets, the interactive features will not work.

To enable widgets:

pip install ipywidgets
#JupyterLab
jupyter labextension install @jupyter-widgets/jupyterlab-manager
#VScode
code --install-extension ms-python.python
code --install-extension ms-toolsai.jupyter
code --install-extension ms-python.vscode-pylance
code --install-extension ms-toolsai.jupyter-keymap
code --install-extension ms-toolsai.jupyter-renderers

For classic Jupyter Notebook, just installing ipywidgets is usually sufficient.

---

References

• HuggingFace Transformers Documentation
• PyTorch Documentation
• dateparser Documentation
• scikit-learn Documentation
• Jupyter Notebook Documentation
• Matplotlib Documentation
• seaborn Documentation
• ipywidgets Documentation

---

Author

OperatorSDG (Sourojeet Dasgupta)