How to pass intent with entity during "interactive" mode

Hi, When running in shell mode in Rasa, I am able to send intent and entity directly as /intent_name{"entity_name": entity_value}. I am trying the same in interactive mode, but the interactive way is not picking up the entity value. It does not give error, but as I need that entity to move further…

Is there a way to add a list of entity values to train for its variation (no lookup)?

Adding regular impressions improved the confidence level… Thanks.

Thank you

Are you suggesting to include multiple regular expressions for that entity as:

nlu:
 - regex: product_id
   examples: |
     - reg_exp_1
     - reg_exp_2

I think this might work, will try.

I have an entity, which is a kind of ID

I know I can provide Lookup Tables, but the list is huge (4M values) so lookup won’t be a good idea. Is it possible to provide a list of samples for each format for an entity?

Can we log each incoming message to Rasa with session id for debugging?

I liked the idea, will read it. But if you have some source like blog or documentation, that would help. Thank you.

–debug is helpful, but it throws a lot of info, I wish there could be a flag or feature to dump only message transactions. Thanks for your reply.

Hello all

We are trying to integrate Rasa with a customer support system. Sometimes we observe a reply from Rasa that we do not anticipate and for us, there is no easy way to find what message was fed to Rasa that caused a specific unexpected response. Is there any way we can force Rasa to collect this information?

Where is trigger_response_selector defined in Rasa demo?

Thank you. So my understanding: we do not need to define intent in nlu, if that intent is passed as-is via another medium or typed directly. Rasa does not complain if the intent is not created with training data, as long as it is in the domain file.

I was looking RasaDemo

In stories such as faq.yml I came across intent trigger_response_selector. I have not found any definition of this intent in any file under data/nlu/. I see action definition as action_trigger_response_selector in actions.py. Can someone please tell me how is this intent defined?

Using two intents in stories with 'or'

Thanks for your quick reply. For the first question, I have used 'or' before. 'or' can be used in two possibilities: Or of multiple intents and/or or of multiple slots. I wanted to try nested 'or'. Example:

# pseudo code
- or:
  - intent: medicine_query
  - intent: information
  - slot_was_set:
    - name: ...

Is it possible to create stories like this:

• or:
• Intent A
• Intent B: conditioned on a slot being filled
• action: action_A

So we want to use Intent A or Intent B (if an intent fills a slot) to initiate a story. Also in a story can we condition on if an entity is detected in the intent. Example:

Rasa-x setup in AWS with HTTP port

First of all, thanks for all your replies. “ngrok http 5005” does not help. I think ngrok can only help I guess, when 5005 is exposed out of Docker containers. At present, no listen port is open on the host for 5005.

If server is down, then the bot can’t do anything

Other than intimating the user that "the server is down, please try after some time later". I am looking for a solution where the user wants to change the form parameters; I am not sure what is a good way to do it. This situation can arise, irrespective of whether the server is up or down.

Handling multiple word entities

Yes I am using DIET and CRF. Ok, I will try to add more examples for DIET to work with multiple word entities. Does the accuracy depend on what language model (word vector model) we are using and if the word vector model has a vector for such 'multiple word entities'?

In domains such as medical

There are many entities that are composed of multiple tokens. In such entities, the meaning of individual tokens can be very different from the whole entity (multiple words). When I run NLU, Rasa recognizes a few of such entities, but in most cases, it splits the entity.

I have a few intents in my setup

Such as affirm, deny, greet, and other work-specific intents. I recently added the capability to handle fallback. Sometimes, I see that nlu_fallback gets a higher match compared to other intents such as deny, affirm, even when I enter the exact same words as example in training data.

Thank you, those links were very useful

I am trying to set up Rasa-X using Docker Compose

I have a few questions. When we link the Rasa-X to a GitHub repository, it gets all the relevant files. Does Rasa-X pull those files, if yes, where does it keep them? I had custom actions in the Git repository, but I think that didn't get included when I pushed to the server.

Hello,

I am very new to Rasa and NLP, spent reading documentation and video and finally have gathered courage to dirty my hands with chatbot creation. I am trying to make a chatbot for a medical shop. This will provide information about medicine and suggest on doses etc based on a person’s profile.

Easiest way to fine-tune word vectors

Hi, I am also trying to create a chatbot in English but in a specific domain with its own keywords and acronyms. I want to use the pre-trained data, as the language is the same, but want to add all new words in the updated model.