Variational Hierarchical User-based Conversation Model

Implementation of Variational Hierarchical User-based Conversation Model (VHUCM) in EMNLP-IJCNLP 2019.

Environment

We run this code in this environment.

• Hardware

  • Intel(R) Xeon(R) CPU E5-2640 v4 @ 2.40GHz
  • GeForce GTX 1080 Ti 11GB

• Software

  • Python 3.7.4
  • PyTorch 1.3.0
  • NumPy 1.16.4
  • CUDA 9.0

Run the code

Training a model

To train the model, run the RunTrain.sh file. To understand the meaning of arguments, please see the config.py file.

Generating responses

To generate responses from trained model, run the RunExportTestSamples.sh file. It outputs txt files that have a set of input conversation context, generated responses, and ground truth response.

Evaluating a model by generated responses

To evaluate the model, run the RunEval.sh file. It outputs the score of BLUE, ROUGE, and the length of responses.

Data

In the paper, we build and use Twitter conversation corpus. We prepared to release the corpus and submitted a part of the data to EMNLP submission site. However, we got comments from other researchers about the privacy issues. We took the comments and prepare the opening the data to the research community such as removing personally identifiable information and taking Institutional Review Board approval.

We use Cornell Movie Dialogs Corpus to show the availability of the implementation.

Reference

• https://github.com/ctr4si/A-Hierarchical-Latent-Structure-for-Variational-Conversation-Modeling
• https://github.com/jiweil/Neural-Dialogue-Generation
• https://github.com/OpenXAIProject/Neural-Conversation-Models

Discussion

Please let us know if you have any questions or requests by issues or email. (First author of the paper page: https://nosyu.github.io/)