Create a separate task type for collecting text labels

[yk]

text labels can be collected throughout the application. these are things like "profanity", "helpful", etc.

They can be found here: https://github.com/LAION-AI/Open-Chat-GPT/blob/main/oasst-shared/oasst_shared/schemas/protocol.py#L232

We also want dedicated task types:

• label_initial_prompt
• label_assistant_reply
• label_prompter_reply

that explicitly ask the user to provide labels to a piece of text in the form of a task.

[andreaskoepf]

The new tasks should be used to collect peer-quality feedback during rating phase, see: https://github.com/LAION-AI/Open-Assistant/issues/37

[olliestanley]

I've had a quick look at adding tasks for this, see draft PR at #381 if any feedback/comments

[fozziethebeat]

How does this differ from /api/v1/text_labels/? I thought that end point was for storing the arbitrary text labels associated with a message?

[olliestanley]

How does this differ from /api/v1/text_labels/? I thought that end point was for storing the arbitrary text labels associated with a message?

My understanding is that although /api/v1/text_labels/ allows storing text labels associated with a message, it doesn't facilitate text labelling as a task, which is what this issue refers to

[fozziethebeat]

Got it. I think this could help #310 and deal with the situation where users skip tasks.

[olliestanley]

Closing this as resolved by #381