Enterprise virtual agents and chatbots are becoming important for customer support. Through developer tools such as Microsoft LUIS, IBM Watson, the Alexa Skills Kit, and Nuance Mix, natural language understanding capabilities are becoming accessible to developers. New techniques such as deep learning are leading to rapid improvements in the technology. All of these factors are making this an exciting time for natural language understanding applications. This workshop brings attendees up-to-date on natural language understanding technology, covering three important topics: 1) the spectrum of different types of natural language applications; 2) the technologies that underlie natural language interaction; and 3) the natural language understanding development process, looking at some open source options. We work hands-on with some of the current development tools. Attendees are welcome to bring their own ideas for applications for discussion.

Participants are expected to bring a laptop for use during this workshop. Laptops will not be provided.

Kaldi is an open source toolkit for research in speech recognition and speech signal processing. Due to its efficient and accurate implementation of many algorithms, this toolkit is leveraged by many organizations that develop speech technology. This tutorial introduces Kaldi in a gentle manner, to improve your level of comfort with the software. This tutorial is aimed toward providing hands-on experience with Kaldi, including setup and training of a basic speech recognizer. The structure of the code and the basics of Kaldi operations are explained, with a general overview of the toolkit capabilities. Kaldi input and output methods, file structure, parallelization, and piping mechanisms are explained. The focus is on getting a hands-on experience with the tools and getting a basic system working during the tutorial.

Participants are expected to have some Linux OS experience and a level of comfort with command line. They are expected to bring either a MacBook or Linux laptop, or have SSH access to a remote Linux machine for use during this workshop. Laptops will not be provided.

This workshop focuses on the practical aspects of designing home assistants, drawing on video clips of users of the technology showing good and bad design practices. Attendees do paper and pencil hands-on design work. Attendees also learn the design principles that underpin the design of sustained natural conversations, which include the following:

• A brief overview of the technology and the underlying speech technologies
• Who uses home assistants and for what?
• How do I present the correct narrative voice and persona for my assistant?
• How do I signpost the features of my user interface?
• How do I keep the user interface stable while making it feel natural?