Python Spacy Pattern- How to tag a word based on another word?

Question

I'm trying to write a pattern that would tag the whole word as unit based on one substring. Here's example:

terms = [{'ent': "UNIT",
         'patterns':[
            [{'lemma':'liter'}]]}]

text = "There were 46 kiloliters of juice available"

I wanna tag 'kiloliters' as Unit based on this pattern. I tried using 'lemma" but it won't work in this case.

Answer 1

You haven't said which model you're using so I'll use en_web_core_sm:

import spacy
from spacy.matcher import Matcher

nlp = spacy.load("en_core_web_sm")
matcher = Matcher(nlp.vocab)
doc = nlp("There were 46 kiloliters of juice available")

The first thing is that none of these have an ent_type of UNIT:

for tok in doc:
    print(f"'{tok}': ent_type: '{tok.ent_type_}', lemma: '{tok.lemma_}'")

'There': ent_type: '', lemma: 'there'
'were': ent_type: '', lemma: 'be'
'46': ent_type: 'CARDINAL', lemma: '46'
'kiloliters': ent_type: '', lemma: 'kiloliter'
'of': ent_type: '', lemma: 'of'
'juice': ent_type: '', lemma: 'juice'
'available': ent_type: '', lemma: 'available'

Also, as you can see, the lemma of kiloliters is kiloliter. This is a bit annoying as you don't want to have to specify milliliters, liters etc. separately. One alternative is to look for a CARDINAL token (which also includes words e.g. "two liters") followed a regex:

doc = nlp("""
          There were 46 kiloliters of juice available.
          I could not drink more than two liters a day.
          I would only give a child 500 milliliters.
          """
)
pattern = [{'ENT_TYPE': 'CARDINAL'},
           {"TEXT": {"REGEX": "^.*(liter)s?$"}}]

matcher.add("unit", [pattern])

matches = matcher(doc, as_spans=True)
for span in matches:
    print(span[-1].text)

Output:

kiloliters
liters
milliliters