search

clulab / processors

Natural Language Processors
https://clulab.github.io/processors/
417 stars 101 forks source link

Merge master and reconcile differences #756

Open kwalcock opened 10 months ago

MihaiSurdeanu commented 10 months ago

Thanks @kwalcock !

kwalcock commented 10 months ago

One test is failing: TestHash. This is because it makes use of events.yml which has

- name: people-eat-food
  priority: ${rulepriority}
  label: Eating
  example: "John eats cake"
  graph: "hybrid"
  pattern: |
    trigger = [lemma=/eat/ & tag=/^V/]
    food:Food = dobj
    person:Person = nsubj

and the Balaur processor doesn't seem to support hybrid graphs. That file is really there for the OdinStarter, so that won't work as intended. The graph type can probably be changed, but other users of processors will probably run into problems with their rules.

FYI @MihaiSurdeanu

MihaiSurdeanu commented 10 months ago

I think we should populate the hybrid dependencies in this case from the enhanced ones that use just syntax. I will push a fix soon.

MihaiSurdeanu commented 10 months ago

One more idea: to avoid changing external code, maybe CluProcessor should be implemented as an alias for BalaurProcessor?

kwalcock commented 10 months ago

See #759 for implementation of the alias idea.

kwalcock commented 10 months ago

This is failing in TestNumericEntityRecognition where there is a timed test at the end. It is taking approximately 5 times longer to complete than it does in master. I'm not yet sure why.

MihaiSurdeanu commented 10 months ago

Hmm... Other than the "fall" changes, that code should not have changed.

kwalcock commented 10 months ago

Both versions should have fall in them. This branch has the changes from master merged. I'll probably need to time CluProcessor vs. BalaurProcessor again.

kwalcock commented 10 months ago

AFAIK this test has never been a problem before. The time limit used to be 20 seconds. I swapped the limit for a println and it showed the following results:

Scala version Time
2.11 0:00:01:10.920
2.12 0:00:00:34.028
2.13 0:00:00:08.969
3.1 0:00:00:09.176

I'll try to find out where it's coming from.