I have all my training data for ‘inform’ intent with a ‘name’ value declared because always that ‘intent’ is detected means that a name is passed but when i do:
My name is Chris
It detects the ‘intent’ but it doesn’t fill the slot, I try to reduce data, to augment it, nothing works,
What could be the mistake here?
I recently had a similar problem. Turned out that the word-vectors were defined in the spacy language model I used (German in that case). Did you check that? (I do not know how many names are represented by standard word-vectors)
if you are using tensorflow_embedding , then ner_crf is already contained in it . just add a few more data , i have worked with these name extraction and it works fine