Hi , in my training data I have following intentent where I want to also extract username entity
When I test my nlu model with exactly the string from my training data I get the entity extracted
however if I just change the name I don’t have the entity extracted
You are using ner_crf to extract custom entity username. It won’t be able to capture all of the new names. Based on the story sometimes it will but if you have similar stories, it won’t perform well.
Try using Spacy’s pre-trained PERSON entity, you will be able to capture more names this way. However, It is very difficult to extract person names as there can be millions.
Sorry thanks, for a while I worked on smth else. I read documentation about entity extraction. In my config.yml file. Can I add new_spacy into pipieline ? Can I have both ner_spacy and ner_crt in my pipeline at the same time ?
Yes, you can have both of them in the same pipeline. Just add them to your
config.yml file and you can use it.