Create new task and schema for dialogue system datasets

[holylovenia]

Relates #53 and an existing dataloader (i.e., COD).

Question for thought: Since Wizard-of-Oz (WOZ) format is quite popular for task-oriented dialogue systems, should we make our schema similar to it? (cc: @SamuelCahyawijaya @sabilmakbar)

[sabilmakbar]

hi @holylovenia, are CoD & xPersona dataloaders included in this revamp, too?

*CoD has been mentioned

[sabilmakbar]

I think there are a few more dataloaders that will use this schema:

1. Issue Create dataset loader for GlobalWoZ V2.0 #83
2. Issue Create dataset loader for IndoSMD #54

Therefore, shall we make this a high-priority task?

[sabilmakbar]

btw @holylovenia, do you happen to know the dataset link (or at least its schema explanation) for this original WoZ paper?

I also read a few other relevant articles/papers worth a look:

1. WoZ Framework (This is the Wiki Article of WoZ Framework to what the WoZ paper mentions -- from the same inventor, Kelley)
2. MultiWoZ
3. Evaluation using MultiWoZ

[holylovenia]

Hi @sabilmakbar, sorry for the late reply. Yes, let's make it a high priority. As far as I know, the original dataset should be this.

[dehanalkautsar]

Hi @holylovenia @sabilmakbar @SamuelCahyawijaya . I've implemented the schema based on the link above (https://huggingface.co/datasets/woz_dialogue). The PR is in here #237 , let me know as soon as possible if there are some mistakes or not because I need this schema to implement #53 , #54 , and #83 . Thanks!