Issue Entity Ruler : detect Entities with several tokens

[CaSSog]

Hello,

I have an issue in using Entity Ruler for a project in the specific domaine of french Agriculture.

I am trying to recognize specific french agricultural organization in different text over the web.

For this I have created a patterns.jsonl file with specific rules that I call before NER. It's working well except for entities with several tokens and where Spacy recognize other label in the entity name.

An example :
One of this organization is "Chambre d'Agriculture" (in english Chamber of Agriculture). In texts This entity is often followed by a localization. For example "Chambre d'Agriculture de Bourgogne" (in english Chamber of Agriculture of Burgundy) or "Chambre d'Agriculture France" (in english French Chamber of Agriculture).
In these case Spacy either recognize only the region as a localization (for example Burgundy) or recognize all the entity as a localization (for example Chamber of Agriculture of Burgundy as a localization).

Here is the rule I have created in my patterns file :

{"pattern": [{"TEXT": {"REGEX": "(?i)^chambre$"}},{"TEXT": {"REGEX": "(?i)^d'agriculture$"}}], "label": "ORG"}

Here is my code :

import spacy
from spacy import displacy
import srsly
import fr_core_news_md

nlp = spacy.load("fr_core_news_md")
patterns = srsly.read_jsonl("patterns.jsonl")
ruler = nlp.add_pipe("entity_ruler", before="ner")

ruler.add_patterns(patterns)

doc = nlp("the text")
##print([(ent.text, ent.label_) for ent in doc.ents])
displacy.serve(doc, style="ent")