In Spacy, how to match a specific entity type, just after a specific word, whatever there is between?

Question

I would like to match text in Spacy with the following pattern:

• If there is the word "dénomination" or "denomination", I want to match the next 'MISC' entity (entity name from Spacy), whatever there is between the two.

for example in:

text=" Some texte about a company, company number: 254455, Dénomination\n (entire name): NAME_OF_THE_COMPANY , \n, some other informations of the... "

I'd like to extract "NAME_OF_COMPANY" which is recognize by Spacy as the entity MISC

To get the entities with Spacy I do:

for txt in text_file:
    doc = nlp(txt)
    for token_french in doc_french:
    for ent in doc.ents:
         print(ent.label_, ent.text)

But then I tried many pattern as the one below but without any success:

    matcher=Matcher(nlp.vocab)
    pattern = [{"REGEX" : "[D|d][é|e]nomination\s{0,}"},{"REGEX" : "[A-Za-z\n\r\s:)]{1,}"},{"ENT_TYPE" : "MISC"}]
    matcher.add('company_name', None, pattern)
    matches = matcher(doc)

Answer 1

A few things to keep in mind:

• Each dict in the pattern corresponds to one token without surrounding whitespace.

• You can match any number of intervening tokens with {"OP": "*"}.

• It's useful to use validate=True with Matcher() to get more feedback when you're working on new patterns.

I think your pattern might look more like:

pattern = [{"LOWER": {"REGEX" : "d[é|e]nomination"}}, {"OP": "*"}, {"ENT_TYPE": "MISC"}]

The Matcher looks at the whole document, so if you have a long document this will provide not only the next MISC but a match with "denomination" followed by every following MISC. You'd have to select the shortest match from the results separately.