search

nschneid / amr-hackathon

Abstract Meaning Representation (AMR) Hackathon
MIT License
28 stars 12 forks source link

NUM should include fractions and possibly time #10

Closed danielhers closed 6 years ago

danielhers commented 7 years ago

Currently the NUM regex matches decimal numbers, but e.g. 1/7 is parsed as STR. Same for decimals without a leading digit: .270. These should be considered numbers too. Possibly, also times like 15:00 should be NUM, and maybe also ordinals like 3rd and #8722.

nschneid commented 7 years ago

NUM is a syntactic category, not a semantic one—i.e., anything in quotes in the AMR is considered a string insofar as AMR syntax is concerned. Do any of the above expressions ever occur without quotes?

danielhers commented 7 years ago

Good point. No, I think they always have quotes. I guess it's an artefact of the annotation system...

On Wed, Jul 5, 2017, 18:39 Nathan Schneider notifications@github.com wrote:

NUM is a syntactic category, not a semantic one—i.e., anything in quotes in the AMR is considered a string insofar as AMR syntax is concerned. Do any of the above expressions ever occur without quotes?

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/nschneid/amr-hackathon/issues/10#issuecomment-313141223, or mute the thread https://github.com/notifications/unsubscribe-auth/AAQEwzLaOhekNz9LzJVyKR_TSuEubNM9ks5sK64VgaJpZM4ON8VF .