Debiasing Item-to-Item Recommendations With Small Annotated Datasets

This is the code for our RecSys '20 paper. Other materials can be found here:

• Full paper (pdf)
• Slides (pdf)

Setup

This assumes that you have a recent Anaconda distribution of Python 3 installed. To install the dependencies,

conda env create -f environment.yml

Then, activate your new environment

conda activate item2item

and get the datasets

python 0_get_datasets.py

Running the demo

To run the command-line demo that allows you to retrieve item-to-item recommendations interactively,

python 1_run_demo.py

Then, follow the prompts

Input (partial) movie title [empty to quit]: toy
option #
0              Toy Story (1995)
1            Toy Story 2 (1999)
2            Toy Story 3 (2010)
3                   Toys (1992)
4       Babes in Toyland (1961)
5           Toy Soldiers (1991)
6               Toy, The (1982)
7            Toy Story 4 (2019)
8       Babes in Toyland (1934)
9    Toy Story of Terror (2013)

Input option (0-10) [empty to exit]: 0
Recommendations for  Toy Story (1995)
             DebiasedModel                                                      ItemKNN
                     title     score                                              title     score
0       Toy Story 2 (1999) -1.319431                                 Toy Story 2 (1999)  0.632260
1       Toy Story 3 (2010) -1.382858         Willy Wonka & the Chocolate Factory (1971)  0.554588
2      Finding Nemo (2003) -1.532166                          Back to the Future (1985)  0.547485
3  Incredibles, The (2004) -1.544819                              Monsters, Inc. (2001)  0.542195
4    Monsters, Inc. (2001) -1.571283                              Lion King, The (1994)  0.541657
5             Shrek (2001) -1.627429                               Bug's Life, A (1998)  0.538624
6           Shrek 2 (2004) -1.628034               Independence Day (a.k.a. ID4) (1996)  0.535614
7     Bug's Life, A (1998) -1.665477          Star Wars: Episode IV - A New Hope (1977)  0.535263
8       Ratatouille (2007) -1.672807                                     Aladdin (1992)  0.534045
9                Up (2009) -1.722887  Star Wars: Episode VI - Return of the Jedi (1983)  0.532928

Running the baselines

First, fit the models and pick the best on the validation set:

python 2a_find_best.py

Then, get the test set performances:

python 2b_eval_on_test.py

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.

When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.