RankingOptimizationApprochtoLLC

Official Code Framework of the paper "A Ranking Optimization Approach to Latent Linear Critiquing in Conversational Recommender Systems"

If you are interested in building up your research on this work, please cite:

@inproceedings{Li:recsys20,
  author    = {Hanze Li and Scott Sanner and Kai Luo and Ga Wu},
  title     = {A Ranking Optimization Approach to Latent Linear Critiquing in Conversational Recommender System},
  booktitle = {Proceedings of the 14th International {ACM} Conference on Recommender Systems {(RecSys-20)}},
  address   = {Online},
  year      = {2020}, 
  url_paper = {https://doi.org/10.1145/3383313.3412240}
}

Author Affiliate

| |

Algorithm Implemented

1. Ranking-based LLC

Dataset

1. Yelp,
2. Beer Advocate,

We don't have rights to release the datasets. Please ask permission from Professor Julian McAuley.

Please refer to the preprocess folder for preprocessing raw datasets steps. Please refer to https://github.com/k9luo/LatentLinearCritiquingforConvRecSys for previous codebase for Latent Liear Critiquing. This repo adapts PLRec implementation from NCE-PLRec repo. The full version of the recommender system could be found in [https://github.com/wuga214/NCE_Projected_LRec]

Keyphrase

Keyphrases we used are not necessarily the best. If you are interested in how we extracted those keyphrases, please refer to the preprocess folder. If you are interested in what keyphrases we extracted, please refer to the data folder.

Packages

We use Gurobi Optimizer to train our LP Objectives, please see usage in website.

Example Commands

Note: change --data_dir and --dataset_name for different dataset, change --topk for different number of top-K (K = 100 by default) rated items during optimization, change --lamb for different regularization parameter in equation(12).

Reproduce Critiquing for Yelp Dataset

UAC:

python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path uac_random.csv --num_users_sampled 100 --critiquing_model_name uac --keyphrase_selection_method random
python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path uac_pop.csv --num_users_sampled 100 --critiquing_model_name uac --keyphrase_selection_method pop
python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path uac_diff.csv --num_users_sampled 100 --critiquing_model_name uac --keyphrase_selection_method diff

BAC:

python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path bac_random.csv --num_users_sampled 100 --critiquing_model_name bac --keyphrase_selection_method random
python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path bac_pop.csv --num_users_sampled 100 --critiquing_model_name bac --keyphrase_selection_method pop
python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path bac_diff.csv --num_users_sampled 100 --critiquing_model_name bac --keyphrase_selection_method diff

LLC_Score:

python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path llcscore_random.csv --num_users_sampled 100 --critiquing_model_name llc_score --keyphrase_selection_method random
python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path llcscore_pop.csv --num_users_sampled 100 --critiquing_model_name llc_score --keyphrase_selection_method pop
python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path llcscore_diff.csv --num_users_sampled 100 --critiquing_model_name llc_score --keyphrase_selection_method diff

LLC_Rank:

python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path llcrank_lamb200_top10_random.csv --num_users_sampled 100 --critiquing_model_name llc_rank --keyphrase_selection_method random --lamb 200 --topk 10
python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path llcrank_lamb200_top10_pop.csv --num_users_sampled 100 --critiquing_model_name llc_rank --keyphrase_selection_method pop --lamb 200 --topk 10
python reproduce_critiquing.py --data_dir "data/yelp/" --dataset_name yelp/ --save_path llcrank_lamb200_top10_diff.csv --num_users_sampled 100 --critiquing_model_name llc_rank --keyphrase_selection_method diff --lamb 200 --topk 10

Hyperparameter Tuning for Yelp Dataset

python tuning_experiment.py --data_dir "data/yelp/" --dataset_name "yelp" --save_path "tuning_lambda/yelp/"