Tuesday, March 16, 2021

EB at virtual ESSV 2021: Modeling ontology-based dialogs for Voice Assistants in the browser

An Viet My Phan

The 32nd edition of the ESSV 2021 “Elektronische Sprachsignalverarbeitung (Electronic Speech Signal Processing)” was organized virtually in March 3-5, 2021. The goal of the conference was to bring together everyone interested in the field of speech technology research and applications.

The ESSV conference covers a wide range of areas in speech processing such as speech recognition and natural language understanding, speech synthesis and natural language generation, measuring, processing, and modeling articulation, and much more. The participants can present their topics in many ways including verbal paper presentations, poster sessions, and demo sessions.

Talking about new trends in voice management, voice assistants such as Amazon Alexa, Google Assistant, and Siri have become popular especially when the usage of touch interfaces is inconvenient or even difficult to interact with, for example while driving cars. However, to develop an in-vehicle dialog system is still a big challenge since voice assistants only support a limited set of dialog phenomena and a conservative framework to integrate external knowledge sources, whereas modern cars have access to connected and even cloud-based knowledge sources.

On this context, our voice experts Thomas Ranzenberger and Christian Hacker discussed a concept for modeling ontology-based dialogs in the browser. The concept of our tool is called Voice Assistant Dialog Modeling Service. The key point of the concept is to provide ontology-based dialog modeling, testing, simulation, and deployment to current on-market voice assistants.

The concept abstracts the complex definition of dialog rules which are part of our ontology by providing a template-based rule editor. This enables the interactive and test-driven creation of human machine dialogs. The underlying rule engine operates on a knowledge base and enables smart dialogs through reasoning.

 
Voice Assistant Dialog Modeling Service – Modeling ontology-based dialogs for VA’s in the browser

amazon alexa cluster

 

EB have regularly presented at the ESSV conferences since 2016 and successfully published this year the 8th paper in line. We are proud to be part of the conference to exchange our knowledge and expertise in the speech technology field.

Download our Tech paper to learn more about how dialog definition, management, and domain information can be packaged in a uniform, ontology-based model that drives a dialog manager.

Learn more about our EB services for Amazon Alexa integration to build a voice-first in-car experience powered by Amazon Alexa

Check out our other blog posts, subscribe to our newsletter, and follow us on LinkedIn, Twitter, YouTube, and Facebook to learn more about EB!!