MoCoRP: Modeling Consistent Relations between Persona and Response for Persona-based Dialogue

This is the official repository of our journal paper.

Dataset

ConvAI2

You can download ConvAI2 dataset through the ParlAI library link.

Download train_self_original.txt and valid_self_original.txt, and place them in the data/convai2 directory.

MPChat

You can download MPChat dataset.

Download mpchat_nrp.json, and place it in the data/mpchat directory.

Note that we split the mpchat_nrp.json dataset into train_mpchat_nrp.json, valid_mpchat_nrp.json, and test_mpchat_nrp.json for training, validation, and testing, respectively.

Dialogue NLI

Additionally, we use Dialogue NLI dataset for training NLI expert.

Download dialogue_nli_train.jsonl, dialogue_nli_dev.jsonl, and dialogue_nli_test.jsonl, and place them in the data/dialogue_nli directory.

Model training and inference

Please check each directory (MoCoRP and MoCoRP LLM) for model training and inference per each task.

Have any questions?

Please contact Kyungro Lee at lkr981147 at gist.gm.ac.kr

Citation

Please cite our work if you found the resources in this repository useful:

bib