I heard earlier this morning that Fred Jelinek passed away last night. Apparently he had been working during the day: a tenacious aspect of Fred that probably has a lot to do with his many successes.
Fred is probably most infamous for the famous "Every time I fire a linguist the performace of the recognizer improves" quote, which Jurafsky+Martin's textbook says is actually supposed to be the more innocuous "Anytime a linguist leaves the group the recognition rate goes up." And in Fred's 2009 ACL Lifetime Achievement Award speech, he basically said that such a thing never happened. I doubt that will have any effect on how much the story is told.
Fred has had a remarkable influence on the field. So much so that I won't attempt to list anything here: you can find all about him all of the internet. Let me just say that the first time I met him, I was intimidated. Not only because he was Fred, but because I knew (and still know) next to nothing about speech, and the conversation inevitably turned to speech. Here's roughly how a segment of our conversation went:
Hal: What new projects are going on these days?
Fred: (Excitedly.) We have a really exciting new speech recognition problem. We're trying to map speech signals directly to fluent text.
Hal: (Really confused.) Isn't that the speech recognition problem?
Fred: (Playing the "teacher role" now.) Normally when you transcribe speech, you end up with a transcrit that includes disfluencies like "uh" and "um" and also false starts [Ed note: like "I went... I went to the um store"].
Hal: So now you want to produce the actual fluent sentence, not the one that was spoken?
Fred: Right.
Apparently (who knew) in speech recognition you try to transcribe disfluencies and are penalized for missing them! We then talked for a while about how they were doing this, and other fun topics.
A few weeks later, I got a voicemail on my home message machine from Fred. That was probably one of the coolest things that have ever happened to me in life. I actually saved it (but subsequently lost it, which saddens me greatly). The content is irrelevant: the point is that Fred -- Fred! -- called me -- me! -- at home! Amazing.
I'm sure that there are lots of other folks who knew Fred better than me, and they can add their own stories in comments if they'd like. Fred was a great asset to the field, and I will certainly miss his physical presense in the future, though his work will doubtless continue to affect the field for years and decades to come.
Another giant laid to rest. His first influence on me was indirect, through some of the linguists who left his group. Over time, I came to sympathize with both their linguistic insights and Fred's empirical bent. That mutual sympathy inspired my interest in balancing the strengths of the rationalist and empiricist approaches to NLP. Consequently, feature engineering became a focus of my work in supervised NLP.
ReplyDeleteOne of my memories of Fred was his tendency to sit near the front of the crowd for a conference talk. You knew he had lost interest if he slowly stood up and walked down the center aisle -- slowly, with his back to you. :)
--Eric
p.s. For more work on making disfluent speech fluent, I recommend the U. of Rochester Ph.D. dissertation of Peter Heeman.
I realize that must be a common experience upon losing a colleague: You associate them with all the things that you've learned from them over the years (some large, some small), and it all comes rushing back once they're gone.
ReplyDeleteI'm so surprised hearing this news because Fred spent much of his last day hosting me (I gave a CLSP talk yesterday related to his earlier work), and he was vibrant and engaged as usual. Over lunch he told Jason and me some of his legendary stories in the War and Communist times that we would never know otherwise. Then we had long technical discussions with his two students about how to revive his old work with my new algorithm, which was one of the problems he was very interested in these days. During my talk (which was at the end of the day) I discussed some of his old and less-known work that I highly admire, but he said he now knew nothing about it, which made the audience and ourselves all laughing jovially. That was some last memory of him at CLSP, and I feel so sad that we lost such an inspiring scientist and amicable friend.
ReplyDeleteIn memoriam, I provide some links to Fred's own words.
ReplyDeleteFrederick Jelinek. Five speculations (and a divertimento) on the themes of H. Bourlard, H. Hermansky, and N. Morgan. Speech Communication, Volume 18, Issue 3, May 1996, Pages 242-246. ((In this paper, Fred talks about the origin of the famous 'fire a linguist' quote))
Acceptance Speech for Honorary Doctorate from Charles University
Fred's speech on receiving the ACL lifetime achievement award.
About ten times in the past decade, Fred called me on the phone out of the blue. "Noah, it's Fred." Then he would launch into business. One time in 2001 it was an invitation to Prague, where he was on sabbatical and wanting to talk about linguistics and statistics. Another it was a request to work hard to recruit a particular student. In 2008 it was an urgent plea to submit a workshop proposal. I've never seen anyone who compartmentalized so well: when on technical topics he was direct and didn't waste time or hold punches, but after hours he was beyond charming. It's deeply saddening to realize that we've lost both Fred-at-work and Fred-the-gentleman, and that I've received the last of those unexpected phone calls.
ReplyDeleteAbout 15 years ago, not longer after Fred arrived at JHU, he called me into his office and asked me to spell out the the Lexical-Conceptual Structure for "John usually came home on Saturdays." I was shocked that he wanted to know about deep structures, having known him for years to be a hardcore stats guru, and everyone already knows what he said about linguists....and all that!
ReplyDeleteI launched into a very high level description of the use of GO (for a motion event) in the Spatial Field (Loc), a modifier, and two arguments---leaving out many details. But he said I did it too quickly and vaguely, that I needed to start over, write out every detail on his white board. I was floored. Well, so I did! And he listened intently, especially perking up his ears when I showed that the resulting structure illustrates why divergence mappings are necessary (since, for example, "usually" is the main verb "soler", but it is a simple modifier in English). He was delighted that I had pointed this out and had something to say about it. Wow!
Since that day, despite that I am one of those CS-degree-but-closet-linguist types, we maintained wonderfully collegial and friendly relationship.
Amazing guy.
I'll miss Fred.
Bonnie Dorr
University of Maryland
(just down I-95 from where Fred worked)