Datasets:
What is the difference between contract-nli a and b?
Hi,
Accidentally found this is the only repo for contract-nli on huggingface hub, thanks for the work of pre-processing the dataset and sharing it here. I'm curious about what is the difference between a and b versions. Can't find it anywhere in the original dataset paper. Thank you!
Hi, the one (b version) is the original dataset and has the full-text of the contracts, which means you need to run models that are able to encode up to 4096 tokens. The other one (a version) is a simplified version including only the relevant parts (rationales) of the full-text to resolve the task, in which case you can use BERT-like encoders (up to 512 tokens).
Thank you for the response! This is very helpful. I do have another question: how do you generate the premise for 'neutral' categories in version A?