More generally, imagine a corpus which could identify the conversational function of a sentence, such as, "Have a nice day!", in narrow categories like, "ending a conversation", or broad ones like "small talk", and then produce lists of other sentences in the same categories.

Does anything remotely resembling this exist?

Other interesting categories would be, phrases often used in emails, phrases occuring often in business discourse, different ways to apologize, casual ways to greet people, and so on. Even an encyclopedic document of human-curated lists would be valuable, but an oceanic corpus would be far better.

  • Well, if you can figure out how to separate such sentences computationally, then you can generate such corpora from existing ones. But it'll be hard, expensive, and time-consuming, and you probably won't want to just put your results on the web for free after spending all that time and money on making them. These things do not happen automatically.
    – jlawler
    Mar 29, 2019 at 15:53

2 Answers 2


I'd take a look at 'dialog acts' rather than general purpose discourse functions. This is a good place to start: https://dialogbank.uvt.nl/wp-content/uploads/tdb/2015/12/DialogBank-LRE-v8.pdf

It references the ISO 24617-2 standard which captures intents like Accept Apology, Thanking, Accept Thanking, Init-Goodbye, and much more. The following paper identifies several of the corpora that are available: https://arxiv.org/pdf/1506.00839.pdf

  • Those are really nice resources, thank you! However, while those cover very specific communicative functions, I wonder if there is also a way to amass linguistic data (say, transcriptions of dialogue) relevant to various locations or contexts, such as, a political speech, two friends walking in a park, a business meeting, people on a date, etc.? Apr 6, 2019 at 17:07
  • In addition, is it possible to draw up a list of the many classification systems that exist? Discourse functions, dialog acts, communicative functions... what else is worth knowing about? Thank you Apr 6, 2019 at 17:10

The SPAADIA, OASIS, and Switchboard corpora might interest you.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.