2

I'm looking for a dataset of noun phrases that are special in a particular way. I'm not able to pinpoint the key features of these noun phrases that distinguish them from other ones, however. Here are some examples of what I'm talking about.

Like these:

  • Big Apple
  • machine learning
  • nuclear waste
  • pillow talk
  • blue whale

Not like these:

  • red apple
  • fast car
  • dark night

Questions:

  • Is there a recognized name for the type of thing I'm looking for?
  • Is there an available dataset containing these constructions?
  • Is there a relatively straightforward approach to making such a dataset myself?
Improve this question
2
  • Can you let us know if this is what you're looking for, and either comment on the answers or accept one? Thanks!
    – Adam_G
    May 20, 2016 at 21:50
  • 1
    With the exception of "nuclear waste", your first list are all single words -- specifically compound words composed from two words -- as indicated by having more stress on the first part. Your second list are not single words, but phrases made of two words, as indicated by having more stress on the second part. You can't tell the difference from the spelling, but usually you can tell from the pronunciation.
    – Greg Lee
    Dec 22, 2017 at 0:29

2 Answers 2

Reset to default
1

The term you're looking for is "multi-word expressions." There are tons of datasets for this. One of the newest and most comprehensive is STREUSLE. Check out the page here: http://www.cs.cmu.edu/~ark/LexSem/.

Another goog resource for multi-word expressions is SIGLEX-MWE: http://multiword.sourceforge.net/PHITE.php?sitesig=CONF.

In terms of straightforwardness: NO. This is the "fun" part about them, because the definition of each single word does not compose into the definition of the entire phrase.

Improve this answer
1
1

From your examples it looks like you are looking for terms (as defined in Terminology). You can use several tools for term extraction to find terms in a corpus, or you can look around for terminology databases.

Improve this answer

Your Answer

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

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