Resources for our ACL 2023 SafeConv paper.
SafeConv is a large-scale dataset (160000 prompt-response pairs) with comprehensive annotations for conversational safety:
- binary safety label of the prompt;
- binary safety label of the response;
- unsafe spans in the response;
- safe alternatives for the unsafe responses.
The splited data is in data/. Each line is an instance, which is a dictionary. Below are the meaning of the keys:
| Key | Meaning |
|---|---|
| source | data source; 'P' denotes Pchatbot; 'L' denotes LCCC |
| prompt | dialogue history |
| response | current utterance |
| prompt_label | binary safety label of the prompt |
| response_label | binary safety label of the response |
| unsafe_spans_indices | [start, end] indices of the unsafe spans in the response |
| rewrites | rewritten utterances |
You could cite our paper if you find the dataset is helpful using this BibTeX:
@inproceedings{zhang-etal-2023-safeconv,
title = "{S}afe{C}onv: Explaining and Correcting Conversational Unsafe Behavior",
author = "Zhang, Mian and
Jin, Lifeng and
Song, Linfeng and
Mi, Haitao and
Chen, Wenliang and
Yu, Dong",
booktitle = "Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
month = jul,
year = "2023",
address = "Toronto, Canada",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.acl-long.2",
pages = "22--35",
}