Original Repository: https://github.com/YunYunSeoJANG/BookGNN

0. DATASET OVERVIEW

From Goodreads Book Graph Datasets, includes book/user/review ids, Poetry (36,514 books, 2,734,350 interactions, 154,555 detailed reviews) cf. 'review' consist of 'interactions' where the 'review' text is particularly lengthy.

Below are the examples of the case "book_id": "402128"

1. goodreads_books_poetry

36,514 books

{"isbn": "0151686564", "text_reviews_count": "626", "series": [], "country_code": "US", "language_code": "en-US", "popular_shelves": [{"count": "15419", "name": "to-read"}, {"count": "1762", "name": "poetry"}, {"count": "302", "name": "classics"}, {"count": "187", "name": "favorites"}, {"count": "149", "name": "fiction"}, {"count": "136", "name": "cats"}, {"count": "107", "name": "animals"}, {"count": "104", "name": "owned"}, {"count": "100", "name": "books-i-own"}, {"count": "97", "name": "humor"}, {"count": "68", "name": "childrens"}, {"count": "63", "name": "currently-reading"}, {"count": "63", "name": "children"}, {"count": "55", "name": "classic"}, {"count": "46", "name": "literature"}, {"count": "45", "name": "children-s"}, {"count": "36", "name": "default"}, {"count": "31", "name": "humour"}, {"count": "31", "name": "children-s-books"}, {"count": "29", "name": "poems"}, {"count": "28", "name": "owned-books"}, {"count": "25", "name": "library"}, {"count": "25", "name": "favourites"}, {"count": "24", "name": "english"}, {"count": "23", "name": "british"}, {"count": "21", "name": "art"}, {"count": "20", "name": "childhood"}, {"count": "19", "name": "to-buy"}, {"count": "18", "name": "illustrated"}, {"count": "18", "name": "my-library"}, {"count": "18", "name": "kids"}, {"count": "18", "name": "american"}, {"count": "18", "name": "20th-century"}, {"count": "16", "name": "british-literature"}, {"count": "15", "name": "my-books"}, {"count": "14", "name": "children-s-literature"}, {"count": "12", "name": "picture-books"}, {"count": "12", "name": "home-library"}, {"count": "12", "name": "childrens-books"}, {"count": "12", "name": "childhood-favorites"}, {"count": "12", "name": "young-adult"}, {"count": "12", "name": "edward-gorey"}, {"count": "12", "name": "non-fiction"}, {"count": "11", "name": "read-in-2016"}, {"count": "11", "name": "verse"}, {"count": "11", "name": "poes\u00eda"}, {"count": "11", "name": "fantasy"}, {"count": "11", "name": "read-in-english"}, {"count": "11", "name": "own-it"}, {"count": "11", "name": "nobel"}, {"count": "10", "name": "read-in-2017"}, {"count": "10", "name": "poesia"}, {"count": "10", "name": "poetry-plays"}, {"count": "10", "name": "wish-list"}, {"count": "9", "name": "nobel-prize"}, {"count": "9", "name": "1930s"}, {"count": "9", "name": "re-read"}, {"count": "9", "name": "i-own"}, {"count": "8", "name": "2017-reading-challenge"}, {"count": "8", "name": "2016-reading-challenge"}, {"count": "8", "name": "childhood-books"}, {"count": "8", "name": "funny"}, {"count": "8", "name": "books-we-own"}, {"count": "8", "name": "read-aloud"}, {"count": "8", "name": "nonfiction"}, {"count": "7", "name": "read-in-2014"}, {"count": "7", "name": "to-read-poetry"}, {"count": "7", "name": "childrens-lit"}, {"count": "7", "name": "classic-literature"}, {"count": "7", "name": "other"}, {"count": "7", "name": "mine"}, {"count": "7", "name": "general-fiction"}, {"count": "6", "name": "read-2016"}, {"count": "6", "name": "modern-classics"}, {"count": "6", "name": "nobel-laureates"}, {"count": "6", "name": "cat"}, {"count": "6", "name": "poetry-and-plays"}, {"count": "6", "name": "kids-books"}, {"count": "6", "name": "read-alouds"}, {"count": "6", "name": "short-stories"}, {"count": "6", "name": "nobel-prize-winners"}, {"count": "6", "name": "lit"}, {"count": "6", "name": "want-to-own"}, {"count": "6", "name": "plays"}, {"count": "6", "name": "american-lit"}, {"count": "6", "name": "in-english"}, {"count": "5", "name": "read-2015"}, {"count": "5", "name": "musicals"}, {"count": "5", "name": "home"}, {"count": "5", "name": "talking-animals"}, {"count": "5", "name": "read-in-2012"}, {"count": "5", "name": "have"}, {"count": "5", "name": "adult"}, {"count": "5", "name": "modernism"}, {"count": "5", "name": "want"}, {"count": "5", "name": "read-in-2011"}, {"count": "5", "name": "t-s-eliot"}, {"count": "5", "name": "animal-fiction"}, {"count": "5", "name": "children-s-lit"}, {"count": "5", "name": "england"}], "asin": "", "is_ebook": "false", "average_rating": "4.09", "kindle_asin": "", "similar_books": ["884306", "234", "472443", "51244", "305154", "574889", "201711", "857597", "858497", "864051", "400723", "47564", "1391333", "133380", "285151"], "description": "T. S. Eliot's playful cat poems have delighted readers and cat lovers around the world ever since they were first published in 1939. They were originally composed for his godchildren, with Eliot posing as Old Possum himself, and later inspired the legendary musical Cats.", "format": "Hardcover", "link": "https://www.goodreads.com/book/show/402128.Old_Possum_s_Book_of_Practical_Cats", "authors": [{"author_id": "18540", "role": ""}, {"author_id": "21578", "role": "Illustrator"}], "publisher": "Harcourt Brace & Company", "num_pages": "56", "publication_day": "30", "isbn13": "9780151686568", "publication_month": "8", "edition_information": "Illustrated Edition", "publication_year": "1982", "url": "https://www.goodreads.com/book/show/402128.Old_Possum_s_Book_of_Practical_Cats", "image_url": "https://images.gr-assets.com/books/1327882662m/402128.jpg", "book_id": "402128", "ratings_count": "15716", "work_id": "372536", "title": "Old Possum's Book of Practical Cats", "title_without_series": "Old Possum's Book of Practical Cats"}

2. goodreads_interactions_poetry

2,734,350 interactions

{"user_id": "80d52f5e70f023bd0098ab96599a3530", "book_id": "402128", "review_id": "fbd6a22a155c87a84fba7537f06cc94b", "is_read": true, "rating": 4, "review_text_incomplete": "", "date_added": "Fri Apr 19 08:15:15 -0700 2013", "date_updated": "Fri Apr 19 08:15:15 -0700 2013", "read_at": "", "started_at": ""}

3. goodreads_reviews_poetry

154,555 detail reviews

{"user_id": "3ca7375dba942a760e53b726c472a7dd", "book_id": "402128", "review_id": "28423ff309bc896c071a8d9df4a10e8a", "rating": 5, "review_text": "I have three younger siblings and we grew up watching the musical Cats. We knew all the songs and attempted to do the dance moves too. I remember we used to get trouble for jumping off the sofa too. When I found out that Cats was based off of poems, I really wanted to read them. I asked for the book for Christmas one year and I read them all that day. The poems are beautifully written and actually tell stories, whereas some poems are just descriptions. I have no idea how T.S, Eliot came up with so creative and brilliant with something as familiar as the family cat. Eliot is a great writer and I would recommend this book to anyone who is looking for a break from all the intense, sophisticated poems/books they are usually reading. This book is fun and is guaranteed to brighten your day!", "date_added": "Tue Jun 12 08:59:04 -0700 2012", "date_updated": "Fri Jun 15 11:41:12 -0700 2012", "read_at": "", "started_at": "", "n_votes": 0, "n_comments": 0}

---

For non-Colab Users (For Linux, MacOS, Windows users)

1. Dataset 📀

1.1. Download

Download poetry datasets from Goodreads Book Graph Datasets in goodreads folder.

mkdir goodreads
cd goodreads

# Download
wget https://datarepo.eng.ucsd.edu/mcauley_group/gdrive/goodreads/byGenre/goodreads_books_poetry.json.gz
wget https://datarepo.eng.ucsd.edu/mcauley_group/gdrive/goodreads/byGenre/goodreads_interactions_poetry.json.gz
wget https://datarepo.eng.ucsd.edu/mcauley_group/gdrive/goodreads/byGenre/goodreads_reviews_poetry.json.gz # not used now

# unzip
gunzip *.gz

Then the datasets are stored as following:

cd goodreads
    .
    └── goodreads_books_poetry.json
    └── goodreads_interactions_poetry.json
    └── goodreads_reviews_poetry.json

2. Environment Setup ⚙️

For venv users (python==3.10.12 recommended)

python3.10 -m venv .bookgnn
source .bookgnn/bin/activate
pip3 install -r requirements.txt

For conda users (I haven't checked this yet)

conda create -n bookgnn python==3.10.12
conda activate bookgnn
pip3 install -r requirements.txt

3. Preprocess the Dataset 🔥

Preprocess the downloaded datasets. We first only use goodreads_books_poetry.json and goodreads_interactions_poetry.json. The preprocessed results will be saved in the datasets folder.

python3 src/preprocess.py

cd datasets
    .
    └── books_poetry.json
    └── interactions_poetry.json

4. Training 🚀

./scripts/train.sh

If you want parallelized trianing with multiple GPUs, you can add CUDA_VISIBLE_DEVICES=0 condition in train.sh

You can find out the result plot (train/val loss, train/val ROC) in /train_result_plots

4.1. Training with comics & graphic dataset

python3 src/training_comics.py

5. Wandb Parameter Sweeping 🧹

wandb login

wandb sweep src/sweep_config.yaml
wandb agent pljh0906/Prometheus-GNN-Book-Recommendations/<sweep id> --count <count>

6. Show Result 📈

python3 src/show_list.py

7. Visualization 👁️

python3 src/visualize.py

8. Members 👋

장윤서 (Leader) Data Preprocessing, Skeleton Code Revision, Test Code Setting, GNN Seminar

김준석 Parameter Search, Recommendation System Setting, Demo Setting (show_list.py)

김지환 Code Review & Revision (.ipynb to .py), Model Scale-up (Poetry to Comics)

문재원 Code Review, Recommendation System Setting, Website Development

박준형 Graph Visualization, Code Review, Wandb setting (Parameter Sweep, Optimization)

9. References 📄

networkx, Spotify RS