Add korean dataset

[jeonsworld]

In order for open-assistant to work in Korean, we are working on adding a Korean dataset.

Creating a dataset from zero to end is quite difficult. To efficiently add a dataset, we will proceed as follows.

Current progress is as follows.

1. Crawl the dataset
   • A dataset that includes Korean, such as wikihow, and can be used in the form of "instruction-fulfillment"
2. Machine translation
   • Obtain the Korean dataset by machine-translating the en dataset from the OA Dataset List.
   • Machine translation may not be perfect, but I think it can be filtering in the labeling task.
   • In addition, if you have a list of datasets that I can refer to, please share!

These datasets are expected to be used for labeling and learning for RLHF.

[hyunwoongko]

Hi. I am a project lead of EleutherAI polyglot team. FYI, We have many Korean dataset. If you want, we can support it.

[jeonsworld]

Hi, @hyunwoongko

Currently, I am working on converting the public Korean data into a "instruction-fulfillment" format.

I don't know what type of data you can provide, but any data you provide will be of great help. If possible, please let me know what data you can provide!

Thank you!

[CertifiedJoon]

@ontocord If this issue is relevant, I would love to take over from here. Would that be possible?

[camsdixon1]

@CertifiedJoon if you are still interested in taking on this project, I can assign it to you.

[CertifiedJoon]

@camsdixon1 yeah! please do :)