NELL & dates

[Tim McNamara]

I've noticed that NELL seems to have some difficulty interpreting a
date range as a single date:

"July 2-6" http://rtw.ml.cmu.edu/rtw/kbbrowser/date:july_2_6
"1833-38" http://rtw.ml.cmu.edu/rtw/kbbrowser/year:n1833_38

I've been marking those as false. It would be nice if I could say
"you're close, but it's actually a representation of a range", rather
than just saying that NELL's incorrect. Is there anything I can do to
point NELL in the right direction?

[Bryan Kisiel]

Not just yet. We had tentative plans at one point to add a separate
category for date ranges, but even then there is some ambiguity. Does
"2008" refer to a particular instant in time, or is that a year-long range
of dates? If we're talking about the Olympic Games, it might implicitly
refer to only a few weeks of time, or maybe just the non-contiguous days
when they're doing wrestling matches.

Personally, I've been considering ranges to be valid dates on the
assumption that we'll sort and refile them some day when we reconsider how
best to model time ontologically. We're finally getting started on doing
stuff with time this year, but not that far along yet. I think allowing
ranges is also in keeping with allowing things like "sheep" to be an
instance of "animal" or "northern countries of the EU" to be a valid
"country". We're not really doing much with class vs. instance at this
piont.

bki...@cs.cmu.edu

[Tim McNamara]

Okay, I'll change what I'm doing :)