Client-Reactions

This repository provides part of the code for Baseline Classifiers training and evaluation in our paper Understanding Client Reactions in Online Mental Health Counseling. (Our paper has been accepted to ACL2023 as an oral presentation.)

Dependencies

The code is implemented using python 3.8 and PyTorch v1.8.1

Anaconda / Miniconda is recommended to set up this codebase.

Please install dependencies through requirements.txt.

pip install -r requirements.txt

Dataset

For certain reasons, we are currently unable to make our dataset publicly available. If you need access to our dataset, please contact Anqi(lianqi@westlake.edu.cn).

Training

• Domain Adaption(DA). We perform the masked language modeling (MLM) task on all the collected conversations. You can train your own DA model according to this Transformers document. After Domain Adaption training, please place your model in the folder pretrained_model.

• Classifiers for Automatic Label Prediction. We provide a training script: run_dist.sh, please modify the SPEAKER and label_type in this file. Then train the classifier:

  bash run.sh
  

Test

We provide a test script: "mytest.sh". Once you have completed training, please specify the model_dir and label_type in the mytest.sh file, and then run bash mytest.sh for testing.