Hi guys i have trained a nlu, where there are two entities namely list of employess of my company and a list of employess on my client company with two completely different intent. but the training data for two entities are same. the problem am facing is that the nlu recognizes the intent correctly but theres a confusion in entity extraction.(i. e. the nlu recognies the data of my client name as my employee name enitity), so can anyone suggest me where i need to define the condition to pick entity depending upon the intent.
try to recognize just names(which is already quite difficult unless you are using lookup tables with a fixed list) from the natural language input then use a custom action to find which company the person belongs to. Though i think names shouldn’t be the only identifier to make the difference. what happens when you type ‘Amit Singh’(two most common names/surname in India) in both company?
Thanks @souvikg10 for your reply, actually my question is i have trained those two entities with same dataset so both entity will have [Amit Singh] value in lookup what i was asking is that i want the entity to be recognized by the entities that were trained with the utterences not with the entity of another intent. hope you understand my question.
If i remember welll, i don’t think there is a co-relation between intents and entities. Entity extraction follows a different pipeline than intent classification. All entities examples are taken into one dataset with the BILOU flags(if you are using crf-suite) and then a model is fit to detect patterns.
Thanks @souvikg10 for your reply. It helps me a lot.