Text Classification Workbench

Production-oriented text classification repository with two practical pipelines:

• Spam detection from YouTube comments.
• Multi-class email intent/topic classification on the Enron dataset.

The repository combines lightweight, inspectable baselines with notebook-based experimentation so model behavior can be audited and iterated quickly.

Why This Project

Teams that handle support messages, user-generated content, or inbound communications need reliable classification to reduce manual triage and response latency. This repository provides a reproducible foundation for that workflow.

Repository Structure

01 - Spam Detection/
	Config/
	Data/
	Preprocessing/
	Model/
	main.py
	README.md

02 - Email Classification - EnronData/
	configs/
	dataloader/
	dataset/
	model/
	notebooks/
	main.py
	README.md

Notebook_Pipeline/
template.py

Architecture

1. Data loading

• Deterministic local loading from repository-relative paths.

2. Text normalization

• Config-driven preprocessing functions for cleaning, normalization, and token-level filtering.

3. Feature/model experimentation

• Notebook workflows for classical ML and transformer-based modeling.

4. Evaluation

• Standard classification metrics and saved artifacts for comparison.

Quick Start

1) Create environment

python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r "01 - Spam Detection\requirements.txt"

2) Run spam preprocessing pipeline

python "01 - Spam Detection\main.py" --preview-rows 5

3) Run Enron dataset loader

python "02 - Email Classification - EnronData\main.py" --preview-rows 5

Dataset Notes

• YouTube spam comment CSV files are expected under 01 - Spam Detection/Data/youtube_video_comments.
• Enron processed dataset is expected at 02 - Email Classification - EnronData/dataset/preprocessed.csv.

Engineering Notes

• Paths are resolved relative to the repository, not local machine absolute paths.
• Entry points are scriptable from CLI and suitable for CI smoke checks.
• Notebook content is preserved for experimentation and comparison runs.

Roadmap

• Add typed training and inference modules for classical ML baselines.
• Add tokenizer/vectorizer persistence and model versioning.
• Add unit tests for preprocessing and dataloader contracts.

License

This repository is open source under the MIT License.