14:54:03 <RRSAgent> RRSAgent has joined #voiceinteraction
14:54:03 <RRSAgent> logging to https://www.w3.org/2021/03/24-voiceinteraction-irc
14:54:21 <ddahl> meeting: Voice Interaction Community Group
14:54:27 <ddahl> chair: debbie
14:54:33 <ddahl> scribe:ddahl
14:55:55 <ddahl> agenda:https://lists.w3.org/Archives/Public/public-voiceinteraction/2021Mar/0005.html
15:04:44 <ddahl> topic: dirk's walkthrough updates
15:05:00 <ddahl> https://w3c.github.io/voiceinteraction/voice%20interaction%20drafts/paArchitecture-1-1.htm#walkthrough
15:05:36 <ddahl> dirk: does walkthrough match description?
15:06:30 <ddahl> ...this is just the subset that matches part of the text
15:07:08 <ddahl> action: debbie to update text description in Section 4
15:07:48 <ddahl> dirk: reviews Figure 3
15:09:15 <ddahl> dirk: combination of local stuff and remote
15:09:58 <ddahl> ...knowledge graph can be used to find location given GPS
15:12:13 <ddahl> dirk: ASR is sent in parallel to both NLU's (local and remote) (step 6)
15:13:16 <ddahl> jim: what information is given to provider registry?
15:13:34 <ddahl> ...how does the provider registry work?
15:13:48 <ddahl> dirk: doesn't know anything about utterance semantics
15:14:05 <ddahl> ...it just tries to determine the number of registered IPA providers
15:14:14 <ddahl> ...it can use context
15:14:55 <ddahl> ...or user preferences
15:15:26 <ddahl> jim: what do input and output from provider registry look like?
15:17:25 <ddahl> dirk: provider selection service gets the ASR result
15:18:49 <ddahl> debbie: what information is in step 9
15:20:11 <ddahl> dirk: it's always the same, just a query for all providers
15:21:24 <ddahl> ...the provider registry could be more or less smart
15:22:02 <ddahl> jim: after provider selection service gets an ASR result it queries the Provider Registry for all providers
15:22:41 <ddahl> debbie: what about steps 10 and 11?
15:24:28 <ddahl> dirk: for each provider returned from the registry, authentication information is returned
15:25:14 <ddahl> jim: suggest sending NLU to provider selection service instead of ASR result
15:25:28 <ddahl> dirk: or could just send the raw audio
15:26:08 <ddahl> jim: now the Provider Registry sends all the providers, if we use NLU we could be more selective about what providers we return
15:26:30 <ddahl> dirk: this is the provider selection selection strategy
15:26:47 <ddahl> s/selection selection/selection
15:27:27 <ddahl> dirk: this is the minimum and the overall system could be smarter
15:28:33 <ddahl> dirk: then arrow 6 could be after the NLU
15:28:59 <ddahl> dirk: providers could use their own NLU
15:29:42 <ddahl> jim: what's the difference between the orange (Dialog) NLU and the specific NLU's in the IPA providers
15:30:43 <ddahl> jon: what if IPA provider 1 is American, then it could have full complement of components and functionality.
15:31:25 <ddahl> ...we now have text coming from the Dialog box
15:31:58 <ddahl> dirk: could come up with different walkthroughs describing different alternatives
15:32:28 <ddahl> jim: is one of the goals to enable IPA providers to support any input
15:33:05 <ddahl> ...are the inputs in step 13 in a standardized format
15:33:32 <ddahl> dirk: 1 -- raw audio, raw ASR or NLU
15:34:04 <ddahl> jim: suggest restricting to one alternative
15:34:46 <ddahl> debbie: I agree
15:36:59 <ddahl> debbie: maybe rule out NLU as a format
15:37:13 <ddahl> dirk: raw audio would be the most common
15:37:44 <ddahl> ...or text from a chatbot or ASR
15:38:16 <ddahl> jim: future enhancement would be to standardize NLU, but not covered yet
15:39:52 <ddahl> jim: what is the purpose of the NLU in the Dialog box?
15:40:07 <ddahl> dirk: that's the local NLU that processes some of the queries
15:41:08 <ddahl> action: dirk to change arrows to reflect the primary use case
15:41:32 <ddahl> topic: output side of walkthrough
15:41:58 <ddahl> dirk: arrow 1 is the various NLU results
15:42:34 <ddahl> dirk: asks dialog registry for the best dialog
15:42:57 <ddahl> dirk: may need to query for missing slots
15:43:50 <ddahl> ... gets next dialog move
15:44:20 <ddahl> ...NLG comes up with text, doesn't repeat information that's already known to the suer
15:44:24 <ddahl> s/suer/user
15:44:49 <ddahl> jim: we have a distributed semantic processing, some done in orange and some in blue
15:45:23 <ddahl> dirk: yest
15:45:28 <ddahl> s/yest/yes
15:45:40 <ddahl> jim: this might be too confusing
15:46:04 <ddahl> dirk: there are some implementations but not standarized
15:47:13 <ddahl> debbie: how do you find the Dialog that goes with the NLU
15:48:16 <ddahl> ...if the intent is "book flight" then there could be lots of dialog that could handle that intent
15:49:20 <ddahl> ...how do you associate that intent with the appropriate dialog
15:51:39 <ddahl> ...so actually intents and entities come back
15:52:01 <ddahl> ... and that's mapped to dialogs
15:52:44 <ddahl> dirk: that's correct
15:54:00 <ddahl> debbie: when do you actually get your information from the database
15:55:53 <ddahl> dirk: we haven't gotten to that yet, I could add that
15:56:40 <ddahl> dirk: things need to be clearer
16:00:10 <ddahl> action: debbie to pull together list of outstanding comments
16:01:18 <ddahl> dirk: also collected some suggestions from a presentation
16:01:42 <ddahl> ...what should we do to prove that we did what you describe
16:02:27 <ddahl> jim: eventuall there will be conformance tests
16:02:43 <ddahl> s/eventuall/eventually
16:03:15 <ddahl> rrsagent, make logs public
16:03:22 <ddahl> rrsagent, format minutes
16:03:22 <RRSAgent> I have made the request to generate https://www.w3.org/2021/03/24-voiceinteraction-minutes.html ddahl
18:40:02 <RRSAgent> RRSAgent has joined #voiceinteraction
18:40:02 <RRSAgent> logging to https://www.w3.org/2021/03/24-voiceinteraction-irc
18:40:18 <ddahl> present:debbie,dirk,jon,jim
18:40:26 <ddahl> rrsagent, format minutes
18:40:26 <RRSAgent> I have made the request to generate https://www.w3.org/2021/03/24-voiceinteraction-minutes.html ddahl
18:43:28 <ddahl> ddahl has left #voiceinteraction