README.md

metadata

title: Opinion Summarizer
emoji: 🔍
colorFrom: blue
colorTo: green
sdk: docker
pinned: false
license: mit

Opinion Summarizer

An end-to-end NLP workflow that transforms raw Amazon electronics reviews into compact opinion summaries and provides a semantic search experience.

Features

• Semantic Search: Query thousands of reviews using natural language
• Cluster Summaries: View high-level themes extracted from review clusters
• Abstractive Summarization: Uses Google's Pegasus model for generating summaries

How it Works

1. Data Processing: Raw reviews are cleaned and embedded using sentence transformers
2. Clustering: Reviews are grouped by semantic similarity
3. Summarization: Each cluster is summarized using abstractive summarization
4. Search: Query the review corpus using semantic similarity search

Usage

1. Enter a query in natural language (e.g., "battery life of noise cancelling headphones")
2. View the most relevant reviews ranked by similarity
3. Browse cluster summaries to discover common themes

Technical Details

• Embedding Model: sentence-transformers/all-MiniLM-L6-v2
• Summarization Model: google/pegasus-xsum
• Clustering: K-means with PCA dimensionality reduction
• Search: Cosine similarity over embeddings using scikit-learn NearestNeighbors

Project Structure

├── src/components      # Modular data/ML building blocks
├── src/pipelines       # Executable steps (load→embed→cluster→summarise)
├── artifacts/          # Generated assets (clean data, embeddings, etc.)
├── templates/ + static/ # Flask UI
└── app.py             # Flask application entrypoint

Local Development

1. Install dependencies:

   pip install -r requirements.txt
   

2. Generate artifacts (if needed):

   python -m src.pipelines.full_run_pipeline
   

3. Run the app:

   flask --app app run --port 8000