MCMIPL

The implementation of Multiple Choice Questions based Multi-Interest Policy Learning for Conversational Recommendation (WWW 2022).

More descriptions are available via the paper.

The code is partially referred to UNICORN and SCPR.

Environment Settings

python: 3.7.9

pytorch: 1.7.1

dgl: 0.6.0

Data Preparation

We have released the data, including lastfm_start, yelp_star,Amazon-Book and MovieLens.

If you would like to use your own dataset, you can reference the following step.

1. Put the user-item interaction data into the fold /data/<data_name>, you can see the detail in SCPR.
2. Processing data: python graph_init.py --data_name <data_name>
3. Use TransE [OpenKE] to pretrain the graph embeddings. And put the pretrained embeddings under "/tmp/<data_name>/embeds/".

Training

python RL_model.py --data_name <data_name>

Evaluation

python evaluate.py --data_name <data_name> --load_rl_epoch <checkpoint_epoch>

Citation

If the code is used in your research, please star this repo and cite our paper as follows:

@inproceedings{10.1145/3485447.3512088,
  author = {Zhang, Yiming and Wu, Lingfei and Shen, Qi and Pang, Yitong and Wei, Zhihua and Xu, Fangli and Long, Bo and Pei, Jian},
  title = {Multiple Choice Questions Based Multi-Interest Policy Learning for Conversational Recommendation},
  year = {2022},
  isbn = {9781450390965},
  publisher = {Association for Computing Machinery},
  address = {New York, NY, USA},
  url = {https://doi.org/10.1145/3485447.3512088},
  doi = {10.1145/3485447.3512088},
  booktitle = {Proceedings of the ACM Web Conference 2022},
  pages = {2153–2162},
  numpages = {10},
  keywords = {Conversational Recommendation, Reinforcement Learning, Graph Representation Learning},
  location = {Virtual Event, Lyon, France},
  series = {WWW '22}
}