Code and data accompanying the paper Submission 1207: Structural Characterization of Dialogues for Disentanglement

Model

1. download dataset from http://jkk.name/irc-disentanglement/

2. train our model. Example:

   python run.py \
   --output_dir ./experiments/v7_test_5e-6_bs2_50_128/ \
   --train_batch_size 2 \
   --eval_batch_size 2 \
   --do_train --do_test \
   --do_lower_case \
   --task_name bert_v7 \
   --max_seq_length 128 \
   --learning_rate 5e-6
   

Requirements

Python 3.6

Pytorch 1.10.0

CUDA 10.2