- From: Deborah Dahl <Dahl@conversational-Technologies.com>
- Date: Fri, 7 Jun 2019 10:23:41 -0400
- To: <public-voiceinteraction@w3.org>
- Message-ID: <009801d51d3c$9b3913b0$d1ab3b10$@conversational-Technologies.com>
Fyi, I'll be speaking at the Voice Summit conference (https://www.voicesummit.ai/) on the "JSON Representation of Semantic Information" draft (https://w3c.github.io/voiceinteraction/voice%20interaction%20drafts/emmaJSO N.htm). My talk is "Slots? Entities? Concepts: Agreeing on Formats for Voice Application Tools" (Tuesday, July 23). I'd welcome any thoughts or suggestions from the group about points that you think I should include or emphasize. Debbie This is the abstract: Thousands of developers have developed applications with popular voice application tools like the Alexa Skills Kit, Microsoft LUIS and Google DialogFlow. Anyone who has developed applications with more than one platform will notice that a lot of the same concepts are used in the different tools, but with different names. Some tools may talk about "slots," while other tolls have "concepts" or "entities," even though these are all basically the same idea. As another example, the system's level of certainty about the accuracy of the results could be called the "confidence," or the "score." Not only is the vocabulary different in different toolkits, the overall representation of the results is structured differently. Because of these differences between formats, if it turns out to be necessary to change the platform for whatever reason (cost, missing functionality, compatibility with other systems, or lack of vendor support) converting the old application to a new format can be a time-consuming process that wouldn't be necessary if there were a common, cross-platform format. This talk will go over a recent draft proposal for a common natural language result format, called "JSON Representation of Semantic Information," published by the Voice Interaction Community Group of the World Wide Web Consortium. The format is intended to be able to represent the information contained in any of the popular toolkits. In addition, it provides a common way to include more advanced information, such as inputs that include graphical user actions (multimodal inputs). The goal of this talk is to review the proposal with an audience of experienced application developers, and get their feedback on the specific features of the draft, possible new features, and the general value of a common format. We will also discuss obstacles to the adoption of a common format and how they might be addressed. I have many years of experience developing voice applications with a variety of platforms and I am one of the chairs of the Voice Interaction Community Group. I have worked on speech, natural language, and multimodal standards since 1999.
Received on Friday, 7 June 2019 14:24:04 UTC