One entitie in two intents

Rasa Open Source
1

I have one entitie in two intents, but the entiite is only recognized in one intent.

Here it is the NLU extract where de entitie in the intent “quiero_reservar_en” does not work, but in the intent “informa_hotel” work fine.

## intent: quiero_reservar_en
- me gustaria reservar en [Lagomar](hotel)
- quiero reservar en el hotel [lagomar](hotel)
- quisiera reservar en [Lagomar](hotel)
- estoy interesado en reservar en el hotel [lagomar](hotel)
- quiero realizar una reserva en [lagomar](hotel)
- me podes tomar una reserva en [lagomar](hotel)
- se puede realizar una reserva en [lagomar](hotel)
- me gustaria reservar en [Lagosol](hotel)
- quiero reservar en el hotel [lagosol](hotel)
- quisiera reservar en [Lagosol](hotel)
- estoy interesado en reservar en el hotel [lagosol](hotel)
- quiero realizar una reserva en [lagosol](hotel)
- me podes tomar una reserva en [lagosol](hotel)
- se puede realizar una reserva en [lagosol](hotel)
- Es posible realizar una reserva en [lagomar](hotel)
- me comunico para reservar en el hotel [lagomar](hotel)
- le escribo para reservar en [lagomar](hotel) hotel
- me comunico para reservar en el hotel [lagosol](hotel)
- le escribo para reservar en [lagosol](hotel) hotel

## intent: informa_hotel
- [lagomar](hotel)
- [lagosol](hotel)
- [Lagomar](hotel)
- [Lagosol](hotel)
- [lagomar](hotel)

thank you very much

2

So the intent quiero_reservar_en gets detected successfully, but the entity is not found?

3

Thank you very much for your answer. Yes, the intent quiero_reservar_en gets detected successfully but the entity is not found. Even more, if the word “hotel” is before the entity it’s gets detected successfully.

So, why if it’s write like this, Quiero reservar en [lagomar](hotel) hotel the entitie gets detected successfully. But if it’s write like this Quiero reservar en hotel [lagomar](hotel) the entitie is not detected, and this is the natural way of speack.

4

Can you share you pipeline configuration? And have you tried to add more examples with entities?

5

Yes, here it is the pipeline configuration:

pipeline: supervised_embeddings

i’d tried with more examples, i think the problem is that the entitie in located in the end of the sentens wich is the natural way of speaking, so the NLU can’t detected well. there is a way to fix this?

thanks !!

6

mmhh… should not be a problem that the entity is located at the end of the sentence. You could try lookup tables (Training Data Format), that might help. Other than that, more training data should also help. If that is also not working you can also try out passing custom features to the entity extractor (Entity Extraction), but I would first try out lookup tables.

You could also evaluate your NLU model to see when the model makes mistakes, that might help to identify what the problem is (see Evaluating Models).