search

cylnlp / dialogsum

DialogSum: A Real-life Scenario Dialogue Summarization Dataset - Findings of ACL 2021
170 stars 40 forks source link

non-anonymous version #5

Closed ziweiji closed 3 years ago

ziweiji commented 3 years ago

I find your dataset is anonymous which mute the names as Person1... Could you offer a non-anonymous version?

cylnlp commented 3 years ago

Hi @ziweiji. The reason that we add person tags is not to be anonymous, but to uniform the format and to provide referents for summaries. As stated, we collect dialogue data from Dailydialog, DREAM, and MuTual. Dialogues in Dailydialog contain only utterances (split by line) - no names and tags are assigned before each utterance. For dialogues in DREAM and MuTual, some contain tags such as F&M, or W&M, while some don't contain any.

So the answer to your question is that I am sorry I could not, as the original dialogues do not offer names themselves.