Category Archives: Tuesday

SD202: Standards for Virtual Assistants

SUNRISE DISCUSSIONS – Tuesday 25 April 
08:00 – 08:45

The new W3C Voice Interaction Community Group is collecting use cases for voice interaction with virtual assistants. We are exploring topics such as discovery of virtual assistants with specific expertise, standard formats for statistical language models, interoperability for conversational interfaces, and work on dialogue management or “workflow” languages. This session introduces the Voice Interaction Community group and solicits your ideas for future topics.

Presented by: Deborah Dahl

SD203: Are We Close to Really Universal “Speech as a Service”?

SUNRISE DISCUSSIONS – Tuesday 25 April 
08:00 – 08:45

Today it’s possible to build a software application using services from dozens of online resources: databases, media storage, computing power, error monitoring, customer relations, payment services, and far more. Now that we’ve seen the emergence of speech as a service on our cellphones, how close are we to simple, modularized, network-based speech services for recognition, text to speech, identification, verification, and surveys? And what other services would we like to have for our industry?

Presented by: Moshe Yudkowsky

KEYNOTE PANEL – The Way Forward: Keeping Speech in the Conversation

KEYNOTE – Tuesday 25 April 
09:00 – 10:00

A panel of longtime speech technology specialists delivers contrasting views on the role of speech technology in the world of conversational commerce. This highly popular panel starts with the question, “Are you long on speech” as a preferred user interface, then discusses where, how, and why it will coexist, augment, or enhance a plethora of smart user interfaces for digital commerce and self-service.

Presented by: Dan Miller, Jay Wilpon, Brian Garr, William Meisel, Daniel Hong

C201: The Cooperative Principle in Conversation Design

Track C: VIXD – Tuesday 25 April 
10:45 – 11:30

To the surprise of many technologists, verbal cooperation flouts the rules of math and formal logic—the game of conversation is played with a different set of rules. Come find out what they are, and rethink the rules of grammar design and error repair, how to help users know what to say, how to accommodate users’ diverse communication styles, and how to improve the perception of your application and brand in the minds of everyday people who talk and listen to machines.

Presented by: James Giangola

A201: What Is So New About Chatbots?

Track A: BOTS – Tuesday 25 April 
10:45 – 11:30

Chatbots have been hailed as an important new human/machine interface that will replace apps and traditional IVR systems. What is new about today’s chatbots? How are they different from previous chatbots such as 50-year-old Eliza, from VoiceXML dialogue systems, and from embodied conversational agents? Is there a danger that developers of today’s chatbots will ignore lessons learned from the past and are in danger of reinventing the wheel unnecessarily?

Presented by: Michael McTear

B201: Should I Add Voice Interaction to My Hardware Product?

Track B: TALKING WITH THINGS – Tuesday 25 April 
10:45 – 11:30

Hardware companies face challenges when implementing speechbased services into their products. Challenges include hardware design considerations such as microphone placement, acoustic dampening materials, and digital signal processing components, as well as considerations around trigger word implementation, branding, and the use of multiple voice services. This presentation explores the trade-offs and challenges when adding voice interaction to hardware products.

Presented by: Leor Grebler

D201: Are SLMs the New Panacea?

Track D: STATISTICAL LANGUAGE MODELS – Tuesday 25 April 
10:45 – 11:30

Statistical language model (SLM) grammars have been shown to have a positive effect on customer experience. However, there is often a lack of appreciation of complexities surrounding the building, deployment, and support of SML. This session examines different scenarios and suggests criteria to help gauge whether an application needs an SLM-type solution or not. Additionally, a comparison is presented of the performance of SLM grammars vs. rule-based grammars on some of the most challenging recognition tasks.

Presented by: Dmitry Sityaev

C202: In Conversation, There Are No Errors

Track C: VIXD – Tuesday 25 April 
11:45 – 12:30

When interacting with a virtual assistant, the centerpiece of the user’s experience is the conversation itself. This means that each “error” is an opportunity for the designer to forge a meaningful exchange between the virtual assistant and user. Let’s leverage users’ mental models of how everyday conversations unfold in the negotiation of meaning. Learn how to frame a new way of approaching conversation design, in which so-called errors become organic turns in the dialogue—moving conversational design forward naturally.

Presented by: Nandini Stocker