Eric, very interesting approach. If it works, it might also mean that Watson would be able to automatically recognize the data for the "Which location" step from the conversation even when a synonym is used, without having to train the AI much for that step (maybe a little adjustment in the beginning to help Watson figure things out). This may also present the advantages of preserving the context and history of the conversation. It is definitely worth exploring.
------------------------------
Antonel Neculai
------------------------------
Original Message:
Sent: Fri April 21, 2023 01:29 PM
From: Eric Wayne
Subject: The new Watson Assistant
One way we often see customers handle this pattern would be to define an Action such as "Request directions" and then define a "Which location" step that would request the location for which they are asking directions. In the step definition you would specify how to collect the response. For example, you could use the options response type. Then when a user asks "How do I find the bathroom," the "Request directions" action will start, and it can automatically fill in the data requested for the "Which location" step, skip that step and go on to the step that is conditioned on the value of the location collected. In this way, the Action includes the intent, entity and dialog definition that you would have to create separately in the Classic Dialog way of building.
Thanks,
Eric
------------------------------
Eric Wayne
Original Message:
Sent: Tue November 15, 2022 01:51 AM
From: Graham Walker
Subject: The new Watson Assistant
Hi Daniel,
I hadn't seen the article referred to by Torben and read it with interest. My personal experience is somewhat less glowing than Niklas' "awsome" experience. That said I really like the new Action ability to have individual API callouts wherever needed rather than having to use a single webhook (or on a larger scale an entire orchestration layer). What I find missing in Actions is the concept of Entities. I use Entities fairly extensively. I use an Intent to determine the overall goal of the utterance but then I tend to use Entities to refine specifics of what has been said. Eg. How do I find the bathroom, where is the kitchen, how do I get to the bedroom. Each of these utterances would be used to build a "#location" Intent and then I would use Entities to refine the specific location the user is interested in. Without Entities, how do you do this in Actions?
Regards,
Graham
------------------------------
Graham Walker
Original Message:
Sent: Mon November 14, 2022 02:25 PM
From: Daniel Toczala
Subject: The new Watson Assistant
Torben,
You can choose which "experience" you would like to use when you are in the tool. You can use the "old" experience with Dialog nodes, or the "new" experience with Actions. When using the API, you can use both the V1 and the V2 API with your Watson Assistant instance. You just need to decide how you would rather use API.
So to answer the original question, it is possible to deploy a "old" version of your bot, and then create asecond "new" version of your bot as a second assistant. They would be deployed to different endpoints - but both could be deployed from the same Watson Assistant instance on the IBM Cloud.
------------------------------
Daniel Toczala
Community Leader and Customer Success Manager - Watson
dtoczala@us.ibm.com
Original Message:
Sent: Fri November 11, 2022 06:22 AM
From: Torben Achilles
Subject: The new Watson Assistant
Hi, WA Experts,
I have a question regarding the new Watson Assistant from the following post. http://heidloff.net/article/the-new-watson-assistant-is-awesome/
Is the changeover to the new WA a complete changeover of all excisting bots? Or is there a possibility to test this in advance on one bot without converting all of them directly?
Thanks in advanced.
------------------------------
Torben Achilles
------------------------------
#WatsonAssistant