Modified hatexplain dataset generation

adaamko / POTATO

XAI based human-in-the-loop framework for automatic rule-learning.

MIT License

47 stars 8 forks source link

Closed GKingA closed 2 years ago

GKingA commented 2 years ago

Changes:

read_hatexplain modified extensively to generate a more diverse array of files
explainable_sample and explainable_dataset now contain a new field called rationale_id, which is the list of token indices, that have been tagged as rationale. The explainable sample also expects a potato graph upon initialization.