AES Los Angeles 2016
Product Development Track Event PDSS3
Saturday, October 1, 2:00 pm — 4:00 pm (Rm 402AB)
Product Development Super Saturday: PDSS3 - Speak2Me Demo and Natural Voice Input
Presenters:Paul Beckmann, DSP Concepts, Inc. - Santa Clara, CA USA
Dan Carter
Mark Trainer, Minimum Phase LLC - Northridge, CA, USA
Abstract:
AES Super Session: How to Develop a Killer Audio Product in One Day!
Following our lunch break, the PDSS Team will host a Demo to show off the Speak2Me product to PDSS attendees in a “mini” trade show floor co-located in the PDSS room. You will see and hear how it performs from a variety of consumer uses. The development team will be available to answer questions from PDSS attendees.
Voice Input using Natural Language Understanding is probably the hottest feature set in consumer audio products today from loudspeakers to remote controls to doorbells to assistive living devices and beyond. With the growth in IoT devices predicted in the billions, a key input to many of these devices is listening and processing information from humans. Barry Roitblat of VoiceBox will lead a discussion of the primary considerations for creating a great user experience using natural language understanding.
Natural Language Understanding is about more than just voice recognition. Once the system understands the words spoken, it must still discern the meaning and then take the appropriate action. To do this, requires Voice AI. Voice AI combines a number of features and techniques, including:
• Context – historical, conversations, environmental, or personal information that lend to the understanding of the current request.
• Semantic Understanding - parsing, knowledge, and reasoning to determine the meaning of a request from the language structure, and infer appropriate queries and actions to fulfill the request
• Machine Learning - adapt to different dialects and pronunciations, understand new words and aliases, derive new relationships, and learn new phrases or other language forms from previous interactions
• Dialog – use natural language responses and multi-modal interaction (speech, touch, gesture, etc.) to request additional information needed to fulfill a request