Saras Institute Interviews Talkin, Jelinek, and Stevens

BY FIROJ ALAM, SUNAYANA SITARAM, NAVEEN PARIHAR, & SAMER AL MOUBAYED

We continue the series of excerpts of interviews from the History of Speech and Language Technology Project. In these segments David Talkin, Frederick Jelinek, and Kenneth Stevens discuss how they became involved with the field of speech and language technology. These interviews were conducted by Dr. Janet Baker in 2005 and are being transcribed by members of ISCA-SAC as described previously and last newsletter.

Firoj Alam (BRAC University, Bangladesh) transcribed the David Talkin excerpt
Sunayana Sitaram (National Institute of Technology, Surat, India) and Naveen Parihar (Mississippi State University) transcribed the Frederick Jelinek excerpt
Samer Al Moubayed (Royal Institute of Technology, Sweden) transcribed the Kenneth Stevens excerpt
Syxtus Gaál is coordinating the transcription efforts

David Talkin

Q: One question that we're asking everyone is how did you get into speech?

A: It's hard to say exactly where to start, but I guess I'll start when I came and joined my father in Washington D.C. in about 1964, and he encouraged me to go back to high school, among other things. But he also, at the time, was working on a spectrum analyzer for a friend of his who was a violinist and who was a collector of violins. So my dad was trying to lash up some kind of a plug-in, a Tektronix spectrum analyzer plug-in to some other chromeo-oscilloscopes lying around to make a spectrum analyzer and I get involved that project. As part of that we were modifying a television set to be the display device and then I started tinkering with television sets and hooking them up to stereo systems and putting in band-pass filters and things like that to make artistic designs on them.

Q: About how old were you, when you were doing this?

A: I was eighteen/nineteen.

Q: Cool...

A: And finally my high school history teacher, in this last year of high school I was finishing, got wind of this and he was involved in a night club scene. His day job was being a very civilized and good history teacher, and his night job was being a maitre d' at a couple of clubs around town. And he thought it'd be cool if I would install one of these modified television sets in one of the clubs. So we did that in fact, we installed them in two or three of the clubs... and oh, I don't know if it was a big hit or not.

Q: What did it display?

A: Well there were several variations on it. This very simplest was simply a matter of letting the thing run in its normal horizontal/vertical sweep ray and modulating the R, G and B guns with band passes of music! So what you ended up with were these sort of sheets of waves of color overlapping, and mixing and so forth; and semi-synchronizing, that sort of thing. Anyhow, it was proper to the time, the late sixties.

Q: So where does it go from there?

A: So it then turned out that my history teacher got a job at Gallaudet College in Washington DC, teaching history there, and he came out to know Mac Pickett, James Pickett who was running the lab there, at the time they were interested in visual transformations for the hearing impaired. So they were looking at ways of getting a visual image of speech patterns and he thought that there might be some common interest. So he encouraged me to drop by Mac's lab and bring some of my stuff with me. And Mac was mildly amused, he was sort of winding the visual aids project. By the way I have done all the stuff up to that point using vacuum tubes. This was in the mid- to late sixties, and of course most everybody else was doing everything with transistors, but my father was still very sceptical of transistors. He had a huge amount of tube gear lying around his apartment. And here is Mac's lab is just, you know, completely outfitted with pretty nice modern up-to-date scopes, and transistor gear and everything else, and I just asked if I could hang out, see what I could learn. So initially I started working there for free, and I basically served as a technician for a while, and they let me use the lab for my artistic stuff. And then after about two or three weeks they decided to put me on payroll. Anyhow... that was my longest period of employment at one place, 11 years. I started there in, I guess, '68 until I left in '79.

Q: Now, where did you go to college?

A: Man, I went to college in lot of places. I finally graduated from Johns Hopkins in Baltimore, in fact. I thought I knew everything there was to know about electrical engineering from hanging around Mac's lab for few years, so when I enrolled to go back to school at Catholic University in about '77 -'78, I took mechanical engineering. 'Cause I had been working with Ingo Titze on the mechanical modeling of vocal folds. Of course he was a physicist, but he was very interested in the mechanical properties of vocal folds and identifying the element model of the folds using computer simulation. I was serving as his research assistant during that period. And I got interested in the materials and things like that, so I thought I'd go back to Catholic. So I started that, and then I got an opportunity to work in a lab at Johns Hopkins in one of the medical clinics and as part of that I would get free tuition to Johns Hopkins. The catch was that in order to get free tuition I had to be a full time employee. The other catch was that in order to be a matriculating student I had to go full time (laughs).

Q: (laughs) That just means you don't sleep!

A: I got through it but I think one more semester and I would have been even more psychotic than I usually was.

Frederick Jelinek

Q: First, and maybe the only question that we're asking everyone is, how did you get into speech?

A: Well this is a long story... Well, originally when my wife came to the United States, she got a fellowship arranged by Roman Jakobson to MIT and started taking courses with Chomsky. And when this was the case, I decided or I was attracted to that field and said that maybe I should switch from Electrical Engineering to Linguistics. And my advisor at that time was Fano and he got very upset with the suggestion, he said that I have to finish first and then maybe I can switch.

So, I finished first and then I went interviewing to various universities to become an assistant professor and one university where I was, it was Cornell. At Cornell a famous linguist by the name of Hockett, Charles Hockett came to my talk and said that he is always interested in doing linguistics from the point of view of information theory; my talk was on information theory, and that was my thesis. So, I then accepted the offer from Cornell and came to Cornell and nothing happened then. Charles Hockett was nowhere. And so after about a month or two I thought he would contact me, I went and contacted him, and he said he is no longer interested in linguistics - he's writing operas!

Q: (laughs) That's great!

A: So in 1968 was my sabbatical and I spent that sabbatical at IBM and they still didn't do any speech work, what they did is information theory, communications etc. But in the summer of '72, I ran out of support money for the summer, so I called my friends at IBM, and they said well come on over, we're going to do speech recognition. And so I went there for the summer.

Q: And who was there at that point?

A: Well, at that point the person who was supposed to be in charge was Raviv, and they were just assembling the team, and others they got from Raleigh. They got Dixon and Tappert, and Bakis was there, I don't know who else, Peter Franaszek... I probably could remember other people. So they established this group, and they established it with some linguists. So Stan Patrick kind of joined it, and her name was Robinson, I forget what her first name was... and Fred Amaro. So there were linguists there etc. And then Raviv got an offer to be the head of the science center in Tel Aviv, or in Haifa. And so they started looking for people and somehow they asked me, would I be willing to be in charge.

Q: So when was that?

A: That was July '72, so I came there in June and in July there was this thing. And I called it to my wife and my wife always wanted to be in New York rather than in Ithaca so she said, "Take it". So I took it! So I took a year leave of absence and then I took another year leave of absence and then they wouldn't give me a third one, so I resigned from Cornell and came there. So the point is that I was always interested in speech or language for ten or, I don't know, for eight years before, but this was kind of an opportunity to get in.

Kenneth Stevens

Q: We are here in the library of Professor Kenneth Stevens at MIT. I guess one of the questions that we have been asking everybody is: how did you get into this game [laugh], this field?

A: It was almost by chance actually. I came to MIT as a graduate student, and Leo Beranek, whom you know, was needing a teaching assistant in his course - he was sort of a new professor here - in Acoustics. And he noticed that I took once in Toronto, where I went to school, a course in Acoustics, and he asked if I would be his teaching assistant. And I was, and it turned out that there is a new project on speech in the acoustics lab, building 20. So he asked if I would be interested in working on that, and I said fine, it had to do with speech intelligibility, and that basically got me into the field. It wasn't too long after that, that Gunnar Fant came as a visitor to the lab and stayed for a couple of years at the Acoustics lab, and that definitely got me interested in the acoustics part of the field. So it's sort of by chance that this happened.

Q: So you ended up doing your doctorate in...?

A: In Speech Perception.