Semantic Book Recommender (NLP)

End-to-end semantic book recommendation system using embeddings + vector search, with category and emotion-based filtering in a Gradio dashboard.

What it does

• Takes a natural language query (example: "mystery thriller with detective")
• Retrieves similar books using semantic embeddings + vector search
• Supports category filtering and emotion-based sorting
• Displays results in an interactive Gradio dashboard

Demo

Tech used

• Python, Pandas
• OpenAI embeddings + ChromaDB
• Hugging Face Transformers
• Gradio

Project steps (notebooks)

1. Data exploration: 01_data_exploration.ipynb
2. Vector search: 02_vector_search.ipynb
3. Text classification: 03_text_classification.ipynb
4. Emotion analysis: 04_emotion_analysis.ipynb
5. Gradio dashboard: 05_gradio_dashboard.ipynb

Results

• Cleaned dataset from 6,810 records to 5,197 high-quality entries.

How to run

This repo currently contains the full pipeline as Jupyter notebooks. Packaging into a single runnable app (requirements.txt + app.py) is in progress.

Report

See Final Report.pdf