Getting to Know Your Fellow Researchers: Shrikanth Narayanan

What sparked your interest in speech and language processing?

It’s the essential human centered nature of the field that attracted me to it--that speech is a signal produced for humans by humans, and that it offers tremendous opportunities that naturally blend science and engineering in both the methods and in the design of possible technology applications.   I was also lucky to be exposed to the field early in my graduate school days at AT&T Bell Labs; I felt like a kid in a candy shop to have a chance to learn about everything from basic speech acoustics to the design of conversational dialog systems (and everything in between) there.

How do you think speech and language processing is changing the society for the new generation?

Although speech technologies have been in the real world so to speak for a long while (e.g., speech coders), the sheer number of speech-centric applications deployed and used everywhere today—in phones, in cars, at home, etc.—have connected speech and language technologies more directly to the broad society. But I believe that these have merely scratched the surface of possibilities. I am especially excited about the rich speech and language analytics that can illuminate the human condition (including when any of the underlying systems breakdown e.g., at the neuro-cognitive-motor levels) and can enable a whole range of applications notably in the health realm. This I feel will have even farther reaching impact on the society.

What is your holy grail in speech and language processing? When will we achieve it?

Understanding the mechanisms of how we produce and process speech for me continues to be one of the most fascinating domains for inquiry. This includes uncovering the rich encoding of information in speech regardless of who produces and in what context, and illuminating and dealing with the inherent vast variety, heterogeneity and variability in its decoding. Engineering and scientific advances are helping us inch closer toward this goal, but I think it will be an ongoing challenge at least for a while.

Do you have any specific advice for students, junior faculty or others early in their careers? 

If there are questions and problems that you find interesting and important you should go for them with full heart even if they deviate from the mainstream (and even if deemed to be and tagged “fringy”!). The other incredibly invaluable aspect of my work experience has been the magic of collaboration especially in helping take on challenging problems—of which there are plenty in our field.

What development in the field has most surprised you? Was there a hard problem that turned out to be easy?  An easy problem that proved surprisingly difficult?

The re-entry and rapid pace of sheer domination of neural network methods across the various areas of speech and language processing was indeed surprising.

In terms characterizing the various problems in speech, I view them as variedly challenging and none that easy! But problems that I find continuing to be difficult include creating biologically plausible computational models for speech, as well as untangling (and putting back) the multiple sources and multiple levels of variability contained in speech.

Oftentimes, the work that people get noticed for is not the same as the work which they find most exciting/rewarding/interesting. Which of your publications is your favorite? Why? 

This is a tough question! I really love the stuff I do and the papers that I have had the privilege of contributing to—from those describing experimental and theoretical work to those reporting laborious data collection/data release and engineering systems design. I am particularly fond of my papers on speech science (creating/introducing novel imaging methods for studying the vocal instrument), on emotions (representing some of the early work on emotions in speech and spoken language) and the series of papers on children’s speech and speech technologies for children.