Project discovering creative approaches to recommender systems.
See presentation for evaluation and practical info: simple_book_recommender.pdf.
Using Book-Crossing Dataset.
- Clone the repo
[email protected]:tomas2211/book_recommender.git - Install requrements
pip install -r requirements.txt - Download and unzip the dataset & trained models
./download_data.sh - (Not required - only when training node2vec model
git submodule init && git submodule update)
For a quick demo, run:
python eval_qualitative.py --interactive --formatThe four tested models will be loaded, and you will be able to enter queries (book names).
The kNN model is reachable through a simple API.
TLDR: https://abiding-ripple-272918.ew.r.appspot.com/query?name=book-name
Protip: pass 'format' parameter for a human-readable response: https://abiding-ripple-272918.ew.r.appspot.com/query?format=1&name=book-name