- From: Deborah Dahl <Dahl@conversational-Technologies.com>
- Date: Tue, 26 Jul 2022 11:52:59 -0400
- To: <public-voiceinteraction@w3.org>
I had an action from the last call to review the Architecture and Interfaces documents to check for inconsistencies. I didn't see any major inconsistencies, but I noticed that we used "dialog manager" in figure 2 of the Architecture document and "Dialog Management" in section 5.2 of the interfaces document. While reading over the documents, it occurred to me that it would be helpful to pull out some of the implicit requirements that motivated some of the architectural decisions. For example (from Architecture (https://w3c.github.io/voiceinteraction/voice%20interaction%20drafts/paArchitecture-1-2.htm) Section 3.2.4 " The Dialog Manager selects the best suited input from the available input alternatives for further processing. For this, it should generally expect that the user may switch the goals and thus dialog flows at any time, but must also consider ongoing workflows that must not be interrupted." I see two good requirements there: 1. the architecture must enable the user to switch goals at any time 2. the architecture must enable ongoing workflows to not be interrupted I think we should prepare a requirements document based on these documents because it would clarify the rationale behind some of the architectural decisions. This would be a good topic for discussion during tomorrow's call. Other comments; 1. In section 3.2.4.2 Session -- the document provides some properties of sessions, but doesn't really define a session. How about the following "A session is defined as a set of one or more turns between a user and an IPA that begins when the user starts to interact with the IPA and ends either through normal completion of the goals of the interaction or timeout if the user doesn't respond to a system utterance within a preset time span. This specification does not define how sessions that are interrupted by errors such as loss of connectivity should be handled. IPAs may release resources such as ASR and NLU when a session ends, but are not required to do so." 2. the caption for Figure 3 in the Architecture document should be "walkthrough for the input path of an IPA", not output path. 3. In the Interfaces document (https://w3c.github.io/voiceinteraction/voice%20interaction%20drafts/paArchitecture-1-2.htm) introduction, a lot of material is repeated from the Architecture document. I believe we discussed this, and I think it's still a good idea in general, but we should say that we are repeating this information from the Architecture document for convenience., in order to avoid confusion. However, the third paragraph in the introduction should be omitted, because where it says "this document is a first step in our strategy for IPA standardization", it is actually talking about the other document (the architecture). 4. In section 4.3 of the interfaces document, it seems to me that the value for the service id in the "deliverResponse" method should be "id of the service that was executed", not the service to be executed. 5. Section 5.1.1.2 in the Interfaces document, "Client activation strategy" is the same as the corresponding section in the architecture document (https://w3c.github.io/voiceinteraction/voice%20interaction%20drafts/paInterfaces/paInterfaces.htm#clientactivtionstrategy) It would be better to use a link to the architecture document instead of repeating the section so that this information doesn't get out of synch.
Received on Tuesday, 26 July 2022 15:53:24 UTC