Types for slot

Hi Rasa Team,

This is my very first post to get some clarifications. Before get into that, really I would like to say thank for your efforts and brought this as an open source.

Note - don’t think badly because i have to compare my queries with other tools also

Okay. Here my queries are,

Query 1:

Is there any custom types for slot? I have checked RASA documents and noticed there are classes called custom component and actions. But, that will work based on stores what we have framed. Is there any possibilities to specify slots and types in nlu.md file itself? Because I am stucking with one scenario right now. The scenario is if i want to get restaurent details of multiple locations, then i should mention redundent samples like below , - can you find restaurent at india - can you find restaurent at chennai - can you find restaurent at us

Instead, if it is being like this, then that will make our life easier. For example, - can you find restaurent at (location)


  • india
  • chennai
  • us
  • mexico

Also, noticed there is an another option called lookup. we have to give redundent samples seems (please confirm) even if I use lookup?. so, whenever i will chat/send request like "can you find restaurent at “india” or can you find restaurent at “chennai”, that can able to identify the intent correctly

I am just comparing with Alexa which they providing the feature. I hope you might aware of that. Eventhough letting you know.

For example, in Alexa model json, there is an attribute called custom types. We can mention more than one value under a particular types and also we can simple use that in intent slots. Below are the sample { “name”: “CustomIntent1”, “samples”: [ “my name is {Names}”], “slots”: [ “name”:“Names”, “types”:“NAMES” ] }, “types”: [ { “name”: “NAMES”, “values”: [ { “name”: { “value”: “XYZ” }, “name”: { “value”: “ABC” } }, }

Not required to redundent the same samples for multiple locations/names/whatever.

Query 2:

can we get nlu response wihtout core like as how was it in the previous versions? Along with the above custom types/slots feature

python -m rasa_nlu.server --path “/models/nlu” -c “/data/config_mitie.yml”

Please bare with me if I mentioned anything wrong. We are making an AI with your tools. So, kindly help us to make yet better

Thanks, Anbu

It seems i found the answer for that.

Query 1 - we can use lookup tables (please confirm) Query 2: RasaNLUInterpreter has one function called parse. By use the function we can get response in json.

Anyway thanks. Let me know if my understanding is wrong

Thanks, Anbu

Hi @artechno, you can’t declare slots in NLU data because slots are filled by rasa core. However you can declare entities, and if you have a slot with the same name as an entity, it will get auto-filled.

If you want to get the response from the NLU interpreter, you can indeed use the parse function. You can also use rasa shell nlu.