ucsd-ai4h/covid-dialogue

covid-dialogue's Issues

English model scores

The experimental results on the CovidDialog-Chineese test set referred in this paper >>https://arxiv.org/pdf/2005.05442.pdf in Table 8 are on which dataset? Are these results on the 600 conversations on covid-19 or on the full dataset released (3.4M) released in this paper https://www.aclweb.org/anthology/2020.emnlp-main.743.pdf. As there seems no improvement in the results even after using such a large dataset. Can you please clarify this?

Results in arxiv version >>

Results in emnlp paper >>

Similarly for the English dataset as mentioned in Table 4 of arxiv version what is the dataset used?

Where is the train_Middle.txt file for DialogGPT

I want to ask about how to generate the train_Middle.txt file for DialogGPT in train.py ?
Thanks a lot for your help

English Dialogue Models

Hello,

Great work guys !! I was going through the paper 'On the Generation of Medical Dialogues for COVID-19', which talks about English and Chinese models and training. I am trying to reproduce the paper's output, but in this repo, I can just see the Chinese model and training code.
If possible can you please provide me with the ENGLISH model and training code. It would be a great help.

Thank You
Raj

Where is the BART model ❓

After going through your paper I found that this repository lacks the BART model architecture. Please add this architecture.