2

Are data set requests on-topic?

E.g.:

Subject: Speech act prediction data set

I am looking for data sets containing dialogues labeled with speech acts. I am open as to the definition of a dialogue, e.g. message board posts are good too (1).

Example (FOL, QST, and RES are speech act categories; opening, info, confirm, positive are speech act attributes):

(G) Hi, good afternoon.   0000    FOL(opening)
(T) Hi, good afternoon.   0001    FOL(opening)
(G) %Uh this is tour guide %uh number one And my name is Lynnette.    0002    FOL(opening); FOL(info-opening)
(G) And how do I address you? 0003    QST(opening)
(T) I'm %uh participant number twenty and my name is %uh John.    0004    FOL(info)
(G) John? 0005    QST(confirm-opening)
(G) Hi John.  0006    FOL(opening)
(T) Yup.  0007    RES(confirm)
(T) I'm good. 0008    FOL(opening-positive)

(1) Qadir, Ashequl, and Ellen Riloff. "Classifying sentences as speech acts in message board posts." Proceedings of the Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 2011.

I'm aware of https://opendata.stackexchange.com/

1

You can see what's officially on-topic vs. off-topic in the help center, though it is true that questions can get closed for other reasons. Abstracting away from the character of your example (the computational orientation), I would say that "where do I find" questions are perfectly reasonable, though I agree that one should always spend some quality time with Google first. I wondered if there are any Sorani Kurdish (recorded) speech samples out there and spent enough time to determine that I can't find them, so it would be appropriate to check here. (Appropriate, but I don't think effective, I hate to say; if I really cared, I would email one of the relevant experts).

If you have in mind some kind of restrictions, then of course you should include them in the question – e.g. "in English", "open access", "following Smith's tagging rules...".

3
  • Thanks. In the "What topics can I ask about here?" list I didn't see data requests, however I saw some unclosed data-requests questions such as linguistics.stackexchange.com/q/4237/2680 , so I got confused on what is on-topic. Out of curiosity, what is the issue with the computational orientation of the question?
    – Franck Dernoncourt
    Sep 13 '15 at 19:33
  • @FranckDernoncourt, my point is that computational orientation or not would be orthogonal to the question of on-topicality (since CL/NLP is considered on-topic here), so one could ask about any kind of linguistically-useful resources such as text corpora, speech recordings, grammatical paradigms and so on.
    – user6726
    Sep 13 '15 at 20:47
  • thanks for the clarification.
    – Franck Dernoncourt
    Sep 13 '15 at 20:49
0

They have been allowed so far, but they are almost never good questions. If you can't find what you're after by using search engines then it's unlikely anyone else can either.

2
  • Can't most questions be answered through Google searches?
    – Franck Dernoncourt
    Sep 13 '15 at 2:37
  • @FranckDernoncourt Not really no. While some "what" questions can be answered easily with Google, many good Stack Exchange questions are "why" or "how" questions that ask for an explanation. But "does X exist?" questions are almost never good questions, no matter the content.
    – curiousdannii Mod
    Sep 13 '15 at 3:05

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .