Closed BaiMeiyingxue closed 1 year ago
I think the strategies still work but you may need to manually annotate Chinese dialogue utterances using our strategy schema.
got it, thank you!
BTW, you can refer to our other paper PsyQA. It proposes a Chinese long-text counseling dataset and adopts a similar strategy schema. Hope this helps.
BTW, you can refer to our other paper PsyQA. It proposes a Chinese long-text counseling dataset and adopts a similar strategy schema. Hope this helps. sorry, i just got the message, i'm glad to find the paper!
may i ask for you some detail about the nlp technology? if i wanna realize the ESC language model based on Chinese, in addition to the training data, what other parts of the code do I need to change? could i just use the original pretrained model and parameters? Looking forward to your reply!
You should instead use a Chinese pre-trained dialogue model. Unfortunately, I did not see a direct alternative for BlenderBot (e.g., a huggingface checkpoint) and you should build your own pipeline for data processing and model training.
i got it~ about the paper PsyQA, could i get the source code?
Please follow its repo's README: https://github.com/thu-coai/PsyQA
ok, thank you for your answer!
You should instead use a Chinese pre-trained dialogue model. Unfortunately, I did not see a direct alternative for BlenderBot (e.g., a huggingface checkpoint) and you should build your own pipeline for data processing and model training.
BTW, if we wanna pick one pretrained model, what condition should we think about since we find the Chinese pretrained modelhttps://github.com/thu-coai/CDial-GPT. Whether it can be used?
CDial-GPT is OK but may not be satisfactory. You can consider our released EVA 2.0. It is one of the most powerful Chinese dialogue models open-sourced so far (while there are only a few open-sourced Chinese dialogue models :) ).
CDial-GPT is OK but may not be satisfactory. You can consider our released EVA 2.0. It is one of the most powerful Chinese dialogue models open-sourced so far (while there are only a few open-sourced Chinese dialogue models :) ).
thanks for your effort~
Hi,
Great work! I was wondering whether it is possible to change the English task into a Chinese task. Do you think it is possible to make the strategy come true through the Chinese data and the Chinese model?
Thanks for your reply!