NER custom model with context?

Viewed 19

I'm getting an input sentence like this:

I would like to run func1 with param_1 and param_2

or

Run func2 with param_2 and param_8

Where param_1, param_2 and param_8 are variables (slots) I am aware and I have a NER model that knows how to recognize them (each have finite options I label/annotate and train the model).

Now param_2 is absolutely different in func_2 context, compared to the context in func_1.

How can I pass my NER model the context I need it to learn?

Should I add in my annotations the func name like:

func_1.param_1
func_1.param_2
func_2.param_2
func_2.param_8
etc...

Meaning that in the training process, go to func_1 params and label them as func_1.param_1 and func_1.param_2 and that will give me the "context"?

Please advise what is the best practice to do this?

0 Answers
Related