I have a set of entities that are company codes containing a hyphen. I’ve tried a few different pipelines but whenever the entity is recognised as a synonym the entity returned is the portion of the code to the right of the hyphen.
In my training data I have all the codes listed with synonyms in the format { “synonyms”: [“Netflix, Inc.”, “Netflix Inc”, “Netflix”, “NFLX US”, “NFLX.O”], “value”: “NFLX-US”, “entity”: “company” },
If I post a question regarding ‘Netflix’ the NLU interpreter returns; . . “entities”: [ { “start”: 26, “end”: 27, “value”: “NFLX”, “entity”: “company”, “confidence”: 0.7324651385950932, “extractor”: “ner_crf”, “processors”: [ “ner_synonyms” ] } ]
Why is it not returning “NFLX-US” as the entity as I need the full code for custom actions which will query a db? How can I get the full code returned as the entity?
I have also added a list of all the hyphenated codes in a company.txt (but not the synonym spellings) and have included in my training data as…
“lookup_tables”: [ { “elements”: “data/company.txt”, “name”: “company” } ]