Datasets:

ArXiv:
License:
Dataset Viewer

The viewer is disabled because this dataset repo requires arbitrary Python code execution. Please consider removing the loading script and relying on automated data support (you can use convert_to_parquet from the datasets library). If this is not possible, please open a discussion for direct help.

Dataset Card for FeedbackQA

πŸ“„ Read
πŸ’Ύ Code
πŸ”— Webpage
πŸ’» Demo
πŸ€— Huggingface Dataset
πŸ’¬ Discussions

Dataset Summary

FeedbackQA is a retrieval-based QA dataset that contains interactive feedback from users. It has two parts: the first part contains a conventional RQA dataset, whilst this repo contains the second part, which contains feedback(ratings and natural language explanations) for QA pairs.

Languages

English

Dataset Creation

For each question-answer pair, we collected multiple feedback, each of which consists of a rating, selected from excellent, good, could be improved, bad, and a natural language explanation elaborating on the strengths and/or weaknesses of the answer.

Initial Data Collection and Normalization

We scraped Covid-19-related content from official websites.

Annotations

Who are the annotators?

Crowd-workers

Licensing Information

Apache 2.0

Contributions

McGill-NLP

Downloads last month
472