数据集:
McGill-NLP/feedbackQA
预印本库:
arxiv:2204.03025许可:
apache-2.0? Read ? Code ? Webpage ? Demo ? Huggingface Dataset ? Discussions
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.
English
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 NormalizationWe scraped Covid-19-related content from official websites.
Crowd-workers
Apache 2.0