Closed myedibleenso closed 10 months ago
We use jollyday, which should normalize Thanksgiving Day... Their lexicons indicate that this is covered: https://jollyday.sourceforge.net/names_holiday_default.html
Something is happening as a result of this line: https://github.com/clulab/processors/blob/master/main/src/main/scala/org/clulab/numeric/HolidayNormalizer.scala#L25
@kwalcock : can you please take a look when you can? Maybe we are behind in version numbers?
It looks like that was incorporated as far back as processors 8.4.8. I'll check further.
The problem seems to be that there are two Thanksgiving Days. The second is presumably for Canada. The code only gives a norm if there is a definitive value.
2017-11-23 (Thanksgiving Day)
2017-11-07 (Thanksgiving Day)
I wonder if we can default to the US in case of ambiguities.
On Fri, Nov 18, 2022 at 16:51 Keith Alcock @.***> wrote:
The problem seems to be that there are two Thanksgiving Days. The second is presumably for Canada. The code only gives a norm if there is a definitive value.
2017-11-23 (Thanksgiving Day) 2017-11-07 (Thanksgiving Day)
— Reply to this email directly, view it on GitHub https://github.com/clulab/processors/issues/677#issuecomment-1320667749, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAI75TQSBBH36PHZP4OSX6TWJAJBDANCNFSM6AAAAAASEYO4RY . You are receiving this because you commented.Message ID: @.***>
Are there other ambiguous cases? I wouldn't be surprised (ex Independence Day). Maybe just avoid normalizing if ambiguous or return all assignments?
Was this addressed in https://github.com/clulab/processors/pull/689?
Looks like @kwalcock also updated the artifact registry to use SSL 🎊 ! This will make life easier for those us wrestling with firewalls.
When do you all plan to release 8.5.3?
@myedibleenso, I do not believe #689 was related and we never decided what should be done in ambiguous cases. I would guess ~1 week for new release, but https://artifactory.clulab.org can already be used if you edit the resolver in sbt. Maven might not yet work since it can follow old transitive dependencies.
It turns out not to have been Canada, but Montana instead. At some point in time Thanksgiving was celebrated on a different day there. Canada doesn't come into play because of ManagerParameters.create(HolidayCalendar.UNITED_STATES)
. We don't know anything about the state, so I'm going to try to change the logic so that any state-specific values are not used unless necessary.
This should be taken care of with #771.
Somewhat timely, but I'm running into an issue with the NER for
Thanksgiving Day
: