22otters: A Targeted, Customizable, Intelligent Personal Assistant for Patient Care

Voice-driven Intelligent Personal Assistants (IPAs) such as Siri have become much more prevalent recently. We discuss 22otters, a new targeted, customizable IPA for guiding patients through complex medical procedures, where it has been challenging for Patients to adhere to necessary protocols.

We will discuss specific aspects from the Question Answering (QA) component of 22otters such as:

How specific requirements led to design decisions that differ from more general QA systems
A multi-engine speech recognition strategy
Less focus on automatic knowledge extraction, and more on enabling rapid development of domain and Provider-specific content, and ensuring higher precision and recall
Adding meta-information such as time relative to procedure, which is critical for providing the correct answer
Various ontologies used in different contexts. The tree-based structure of many ontologies was not sufficient; this was augmented with a feature-based approach

Finally we will discuss experiences from deployment of the 22otters IPA in the field.

I am responsible for the Speech Recognition and Natural Language capabilities of the 22otters offerings. I want to use these exciting technologies (which I've been working with since 1985!) to create immensely natural interfaces that Patients and Providers alike can use to get better Healthcare outcomes all-around. I have Bachelor's, Masters', and Ph.D. degrees from MIT, all in Electrical Engineering. I was at Nuance for 13 1/2 years, where as Speech Scientist, Manager, and Director in Professional Services, I built Speech Recognition and Natural Language applications for Fortune 100 clients.

I lead our work on designing and verifying the user experience for all our apps. We have an opportunity to create something special, to empower individuals around many health-related aspects of their lives. I'm thrilled to incorporate speech and language technologies to build easy-to-use apps with a meaningful purpose. For me, our work is also a way to address personal loss of friends to cancer. I've been working in areas related to phonetics and speech recognition ever since getting my PhD in Cognitive Science and Linguistics from UC San Diego. As a Voice UI Designer and UX Researcher at Nuance, I established best practices and created many IVR user-centered designs. At Amazon and NST, I worked on creating multimodal mobile solutions and products.