watsonx Assistant

Is there a way to leverage Watson's disambiguation feature without relying on an assistant (that is, using v1 API).  For example, in my application, I'd like to be able to tell what were the disambiguation options that Watson found, so my application could display them to the user, and proceed from then on, just like an assistant would. This would have advantages over coming up with my own disambiguation algorithm from scratch. (Note: I posted this on the Slack channel too; what are best practices in terms of posts?)

------------------------------
Patricia Goldweic
------------------------------

#WatsonAssistant

No way to do it that I am aware of using the v1 API - not in any auto-magical way.  You could look at the relative confidence for multiple intents, and if the results come back relatively close, you can return and disambiguate those intents that score relatively close (but at this point I think I have you doing your own disambiguation algorithm).

------------------------------
Daniel Toczala
Community Leader and Customer Success Manager - Watson
dtoczala@us.ibm.com
------------------------------

Original Message:
Sent: Tue December 08, 2020 11:14 AM
From: Patricia Goldweic
Subject: disambiguation: can it partially work in a 'non-managed' way?

Is there a way to leverage Watson's disambiguation feature without relying on an assistant (that is, using v1 API).  For example, in my application, I'd like to be able to tell what were the disambiguation options that Watson found, so my application could display them to the user, and proceed from then on, just like an assistant would. This would have advantages over coming up with my own disambiguation algorithm from scratch. (Note: I posted this on the Slack channel too; what are best practices in terms of posts?)

------------------------------
Patricia Goldweic
------------------------------

#WatsonAssistant

Thanks Daniel for replying. Yes, I'm aware of the possibility of doing that, but the harder part would actually be to filter the intents obtained in phase 1 - where phase 1 is as you described it- based on whether a dialog node exists whose condition includes the intent, and that would actually be triggered if the corresponding intent was chosen by the user in phase 1. In other words, so that you're algorithm is able to display as disambiguation options *only* intents that satisfy that, and not any other (that might happen to be recognized with some confidence but would not trigger any existing dialog nodes). This is why I was asking about being able to leverage more of what goes on internally in Watson.

------------------------------
Patricia Goldweic
------------------------------

Original Message:
Sent: Tue December 08, 2020 01:12 PM
From: Daniel Toczala
Subject: disambiguation: can it partially work in a 'non-managed' way?

No way to do it that I am aware of using the v1 API - not in any auto-magical way.  You could look at the relative confidence for multiple intents, and if the results come back relatively close, you can return and disambiguate those intents that score relatively close (but at this point I think I have you doing your own disambiguation algorithm).

------------------------------
Daniel Toczala
Community Leader and Customer Success Manager - Watson
dtoczala@us.ibm.com

Original Message:
Sent: Tue December 08, 2020 11:14 AM
From: Patricia Goldweic
Subject: disambiguation: can it partially work in a 'non-managed' way?

Is there a way to leverage Watson's disambiguation feature without relying on an assistant (that is, using v1 API).  For example, in my application, I'd like to be able to tell what were the disambiguation options that Watson found, so my application could display them to the user, and proceed from then on, just like an assistant would. This would have advantages over coming up with my own disambiguation algorithm from scratch. (Note: I posted this on the Slack channel too; what are best practices in terms of posts?)

------------------------------
Patricia Goldweic
------------------------------
#WatsonAssistant

Look at the suggestions response type. https://cloud.ibm.com/apidocs/assistant/assistant-v2#message

------------------------------
Michelle Miller
------------------------------

Original Message:
Sent: Tue December 08, 2020 01:20 PM
From: Patricia Goldweic
Subject: disambiguation: can it partially work in a 'non-managed' way?

Thanks Daniel for replying. Yes, I'm aware of the possibility of doing that, but the harder part would actually be to filter the intents obtained in phase 1 - where phase 1 is as you described it- based on whether a dialog node exists whose condition includes the intent, and that would actually be triggered if the corresponding intent was chosen by the user in phase 1. In other words, so that you're algorithm is able to display as disambiguation options *only* intents that satisfy that, and not any other (that might happen to be recognized with some confidence but would not trigger any existing dialog nodes). This is why I was asking about being able to leverage more of what goes on internally in Watson.

------------------------------
Patricia Goldweic

Original Message:
Sent: Tue December 08, 2020 01:12 PM
From: Daniel Toczala
Subject: disambiguation: can it partially work in a 'non-managed' way?

No way to do it that I am aware of using the v1 API - not in any auto-magical way.  You could look at the relative confidence for multiple intents, and if the results come back relatively close, you can return and disambiguate those intents that score relatively close (but at this point I think I have you doing your own disambiguation algorithm).

------------------------------
Daniel Toczala
Community Leader and Customer Success Manager - Watson
dtoczala@us.ibm.com

Original Message:
Sent: Tue December 08, 2020 11:14 AM
From: Patricia Goldweic
Subject: disambiguation: can it partially work in a 'non-managed' way?

Is there a way to leverage Watson's disambiguation feature without relying on an assistant (that is, using v1 API).  For example, in my application, I'd like to be able to tell what were the disambiguation options that Watson found, so my application could display them to the user, and proceed from then on, just like an assistant would. This would have advantages over coming up with my own disambiguation algorithm from scratch. (Note: I posted this on the Slack channel too; what are best practices in terms of posts?)

------------------------------
Patricia Goldweic
------------------------------
#WatsonAssistant

Original Message:
Sent: 2/8/2021 2:07:00 PM
From: Michelle Miller
Subject: RE: disambiguation: can it partially work in a 'non-managed' way?

Look at the suggestions response type. https://cloud.ibm.com/apidocs/assistant/assistant-v2#message

------------------------------
Michelle Miller
------------------------------

Original Message:
Sent: Tue December 08, 2020 01:20 PM
From: Patricia Goldweic
Subject: disambiguation: can it partially work in a 'non-managed' way?

Thanks Daniel for replying. Yes, I'm aware of the possibility of doing that, but the harder part would actually be to filter the intents obtained in phase 1 - where phase 1 is as you described it- based on whether a dialog node exists whose condition includes the intent, and that would actually be triggered if the corresponding intent was chosen by the user in phase 1. In other words, so that you're algorithm is able to display as disambiguation options *only* intents that satisfy that, and not any other (that might happen to be recognized with some confidence but would not trigger any existing dialog nodes). This is why I was asking about being able to leverage more of what goes on internally in Watson.

------------------------------
Patricia Goldweic

Original Message:
Sent: Tue December 08, 2020 01:12 PM
From: Daniel Toczala
Subject: disambiguation: can it partially work in a 'non-managed' way?

No way to do it that I am aware of using the v1 API - not in any auto-magical way.  You could look at the relative confidence for multiple intents, and if the results come back relatively close, you can return and disambiguate those intents that score relatively close (but at this point I think I have you doing your own disambiguation algorithm).

------------------------------
Daniel Toczala
Community Leader and Customer Success Manager - Watson
dtoczala@us.ibm.com

Original Message:
Sent: Tue December 08, 2020 11:14 AM
From: Patricia Goldweic
Subject: disambiguation: can it partially work in a 'non-managed' way?

Is there a way to leverage Watson's disambiguation feature without relying on an assistant (that is, using v1 API).  For example, in my application, I'd like to be able to tell what were the disambiguation options that Watson found, so my application could display them to the user, and proceed from then on, just like an assistant would. This would have advantages over coming up with my own disambiguation algorithm from scratch. (Note: I posted this on the Slack channel too; what are best practices in terms of posts?)

------------------------------
Patricia Goldweic
------------------------------
#WatsonAssistant