cooelf / mdfn

Mask-based Decoupling-Fusing Network (AAAI 2021)

Codes accompanying the paper Filling the Gap of Utterance-aware and Speaker-aware Representation for Multi-turn Dialogue

Instruction

Our code is compatible with compatible with python 3.x so for all commands listed below python is python3.

We strongly suggest you to use conda to control the virtual environment.

• Install requirements

pip install -r requirements.txt

• Train the model and predict.

python run_MDFN.py \
--data_dir datasets/mutual \
--model_name_or_path \
google/electra-large-discriminator \
--model_type electra \
--task_name mutual\
--output_dir output_mutual_electra \
--cache_dir cached_models \
--max_seq_length 256 \
--do_train --do_eval \
--train_batch_size 6 \
--eval_batch_size 6 \
--learning_rate 4e-6 \
--num_train_epochs 3 \
--gradient_accumulation_steps 1 \
--local_rank -1 \

Reference

If you use this code please cite our paper:

@inproceedings{liu2021filling,
  title={Filling the Gap of Utterance-aware and Speaker-aware Representation for Multi-turn Dialogue},
  author={Liu, Longxiang and Zhang, Zhuosheng and and Zhao, Hai and Zhou, Xi and Zhou, Xiang},
  booktitle={The Thirty-Fifth AAAI Conference on Artificial Intelligence (AAAI-21)},
  year={2021}
}