Spacy: EntityRuler throws Segmentation Fault in complex pattern case

[imaurer]

How to reproduce the behaviour

import spacy
nlp = spacy.load("en_core_web_lg")
ruler = nlp.create_pipe('entity_ruler')
nlp.add_pipe(ruler, before="ner")
ruler.add_patterns([{"label": "PRODUCT", "pattern": [{"lower": "arial"}, {"lower": "("}, {"lower": "ab987"}, {"lower": ")"}, {"lower": "foo"}, {"lower": "bar"}, {"lower": "baz"}, {"lower": "boo"}]}])
nlp("Arial (ab987) foo bar baz boo")

Results in:Segmentation fault: 11

Your Environment

• Operating System: Mac OS X (10.13.6)
• Python Version Used: 3.6.6
• spaCy Version Used: 2.1.0a1
• Environment Information: built from commit 1759abf1e5703cb568e0f2de56ed1beb0388a6f6 on develop branch

Related to Issue #2513

[ines]

Thanks a lot – also for testing the entity ruler with a real-world use case! This is really helpful. I suspect this case might have something to do with how the entity recognizer handles existing entities in this particular case.

[GregSilverman]

Hi,
I too am getting a segmentation fault on this example (which is a common pattern we are using). (I just installed the nightly build using pip install spacy-nightly)

import spacy
nlp = spacy.load("en_core_web_lg")
ruler = nlp.create_pipe('entity_ruler')
nlp.add_pipe(ruler, before="ner")
ruler.add_patterns([{"label": "Age", "pattern":[{"IS_DIGIT":True},{"LOWER":"years"}]}])
nlp("10 years")

[honnibal]

Thanks for the report! Fixed on develop now. Will be out in the next nightly.

[lock[bot]]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.