This is the source code of the paper "A Cooperative Memory Network for Personalized Task-oriented Dialogue Systems with Incomplete User Profiles", accepted by TheWebConf 2021.
Please contact me by email ([email protected]) if you have any questions.
The original personalized TDS dataset is accessible using this link. And the preprocessed dataset can be found at this link.
Parameters of train.sh and infer.sh are $1_run_node $2_model_name $3_exp_setting $4_other_params.
sh train.sh 104 CTDS _our_t5_pro1_g1_pre1_pdr0.0_k100all_ep250 '--use_profile=1 --use_neighbor=1 --use_preference=1 --profile_dropout_ratio=0.0 --bsl=0 --task=babi-small-t5 --neighbor_policy=k --epoches=250'
sh infer.sh 104 CTDS _our_t5_pro1_g1_pre1_pdr0.0_k100all_ep250 '--use_profile=1 --use_neighbor=1 --use_preference=1 --profile_dropout_ratio=0.0 --bsl=0 --task=babi-small-t5 --neighbor_policy=k --epoches=250'