Wrong entity classification through ner_crf

Hi everyone, I have 2 entities namely “component” and “env”.I have given enough training data for both the entities and it is able to predict the utterences in training data correctly. But if I try to give anything other than utterences it is giving the “component” entity like even if I say hi it is giving me “component”.

Any help will be appreciated.

Pipelines I am using { “name”: “intent_entity_featurizer_regex” }, { “name”: “intent_featurizer_spacy” }, { “name”: “ner_crf” }, { “name”: “ner_synonyms” }, { “name”: “intent_featurizer_count_vectors” }, { “name”: “intent_classifier_tensorflow_embedding”, “intent_tokenization_flag”: true, “intent_split_symbol”: “+” } ]

Hey @rideep. Do you mind sharing a snippet of your training data?

hi @Juste as the data is from my company I cant share it. But how come any utterences can go to an entity if it is not defined there
For example see this {“q”:“hskdhfksjdhi”, “project”: “my_project1”}
“intent”: {
“name”: null,
“confidence”: 0
}, “entities”: [
“start”: 0,
“end”: 12,
“value”: “hskdhfksjdhi”,
“entity”: “component”,
“confidence”: 0.5786415288437734,
“extractor”: “ner_crf”
} ], “intent_ranking”: [],
“text”: “hskdhfksjdhi”,
“project”: “my_project1”,
“model”: “model_20181129-234500”
Even intent is null it is going to component entity.