cooelf / dp_qa_joint_model

Multi-tasking Dialogue Comprehension with Discourse Parsing (PACLIC 35)

DP_QA_joint_model

This is the code for the paper Multi-tasking Dialogue Comprehension with Discourse Parsing

Environment requirement

Python 3.8
pytorch

Instruction

• Train the model and predict.

python run_molweni_mdfn.py \
--model_type bert-molweni-baseline \
--model_name_or_path "bert-large-uncased-whole-word-masking" \
--data_dir "molweni/Molweni/MRC" \
--version_2_with_negative \
--output_dir "experiment_output" \
--do_train \
--do_eval \
--do_lower_case \
--overwrite_output_dir \
--num_train_epochs 2 \
--per_gpu_train_batch_size 8 \
--learning_rate 2e-5 \
--save_steps 500 \
--dropout 0.2 \
--eval_all_checkpoints \
--task 0 \
--weight_decay 0.01 \
--warmup_step 100

The dataset Molweni can be downloaded from https://github.com/HIT-SCIR/Molweni. The parameter 'task' refers to the model to run: 0----multi-tasking model, 1----QA-only model, 2----DP-only model.

Reference

If you use this code please cite our paper:

@inproceedings{he2021multi,
  title={Multi-tasking Dialogue Comprehension with Discourse Parsing},
  author={He, Yuchen and Zhang, Zhuosheng and Zhao, Hai},
  booktitle={Proceedings of the 35th Pacific Asia Conference on Language, Information and Computation},
  pages={69--79},
  year={2021}
}