Obituary for Fred Jelinek at Computational Linguistics
« previous post | next post »
Back on September 15, when I posted the news of Fred Jelinek's death, I promised to say more when I'd had a chance to think about it. Then, a few days later, Robert Dale asked me to write an obituary for Fred to be published in the Computational Linguistics journal. The December 2010 issue is now out, and Fred's obituary is here.
Following Robert's suggestion, I aimed at a broad assessment of Fred's impact on the field, since CL recently published Fred's own detailed account of his professional life ("The Dawn of Statistical ASR and MT", CL 35(4):483-494 , 2009).
John Cowan said,
December 16, 2010 @ 12:33 pm
I see from your obit that John R. Pierce dared to refer to "going to the moon" in the same breath with "turning water into gasoline" (impossible in principle), "extracting gold from the sea" (possible, but insanely difficult and unlikely to be worthwhile), and "curing cancer" (possible and desirable, but still intractable forty years later). In 1969. When a project with the goal of going to the moon was not only underway, but would lead to success that very year.
What an outrage.
[(myl) Yes, I also thought that was ironic. But in fairness to Pierce, who certainly knew about the Apollo Program, what he seems to have in mind is the idea of commercially-viable moon runs, which remains an area where investors would do well to tread carefully.]
John Cowan said,
December 16, 2010 @ 12:39 pm
I also see that Charles Hockett held down a tenure line at Cornell for twenty years (1962-82) while no longer doing anything in linguistic research but instead focusing on composing operas. Any rationally designed institution would have replaced him in 1963 with someone willing to do the work for which he was being paid. And of course Hocketts in 2010 are doing exactly the same thing with the same lack of consequences. Arrgh. My blood pressure can't take it.
Twitter Trackbacks for Language Log » Obituary for Fred Jelinek at Computational Linguistics [upenn.edu] on Topsy.com said,
December 16, 2010 @ 12:42 pm
[…] Language Log » Obituary for Fred Jelinek at Computational Linguistics languagelog.ldc.upenn.edu/nll/?p=2846 – view page – cached Back on September 15, when I posted the news of Fred Jelinek's death, I promised to say more when I'd had a chance to think about it. Then, a few days later, Robert Dale asked me to write an obituary for Fred to be published in the Computational Linguistics journal. The December 2010 issue is now out, and Fred's obituary is here. […]
Jonathan Mayhew said,
December 16, 2010 @ 1:01 pm
What an outrage! Hockett only published seven or eight books after 1963 when he stopped being a linguist, including
1967: The State of the Art. The Haag: Mouton
1973: Man's Place in Nature. New York: McGraw-Hill.
1977: The View From Language. Athens: The University of Georgia Press.
etc..
I guess Rice University was also irrational in hiring him after he retired from Cornell, since he had ceased activity in 1963.
[(myl) I don't think that CL has a budget for fact-checkers, and the text of a Lifetime Achievement Award acceptance speech was presumably not refereed. This is apparently a case where two colleagues had a less-than-satisfactory conversation, which one of them described a half-century later with a bit of poetic license.]
Theo Vosse said,
December 16, 2010 @ 5:08 pm
It's a bit the lion's den here, but where I worked Jelinek was also famous for his quote: Every time I fire a linguist, the performance of our speech recognition system goes up. However, I think that your obituary perfectly places that quote in context.
[(myl) Let me add another bit of context. Fred's favorite adjective for statistical n-gram models (introduced by Shannon, and championed by Fred as a useful approximate model of the prior probability of a sentence) was "moronic". He meant this, I think, as a sincere challenge to himself as well as others to do better — but he honestly felt that doing provably better was hard.]
Bill Benzon said,
December 16, 2010 @ 6:12 pm
This is most interesting, Mark. I studied semantics and knowledge recognition with the late David Hays in the mid-1970s. He was one of the early workers in machine translation (led the RAND effort in the 50s and 60s) and was one of the authors of the ALPAC report. According to Martin Kay (in his speech for accepting the ACL lifetime award) Hays is the one who coined the term "computational linguistics." And, of course, he also advocated that the government spend money on basic research.
While studying with Hays I was also bibliographer for what as then the American Journal of Computational Linguistics (and is now simply the Journal of Computational Linguistics). So I read and abstracted the tech reports on the ARPA Speech Understanding Project as they arrived the the ACJL 'office' (Hays' library). Mere speech recognition seemed like a daunting prospect at the time, as Pierce says. But then, lo and behold, great strides were made within a decade, and with rather different methods from those in the tech reports I had been reading in the 70s. It shouldn't have been possible to achieve those results with those methods. But it was, and Jelinek and others did.
There's no telling what's lurking in a region of the search space that you've not investigated.
I should also note that I do think that one does need an engineering mentality (whatever that is) to figure out how language works (not to mention the mind, or even plain old biology).
Rick Bryan said,
December 16, 2010 @ 6:28 pm
Connecting dots from other fields–
Re tenure: Paul Bracher, on his respected chemistry blog http://blog.chembark.com/2010/12/15/times-up-for-tenure/, considers the costs and benefits of tenure (specifically for chemistry professors) and concludes "Time's up".
Re firing the linguist: Peter Norvig, now Google's Director of Research, lectured to software developers on "Theorizing from Data" (all 53 minutes are at http://www.youtube.com/watch?v=nU8DcBF-qo4). At 26:45 he gets to the part about how Google's probabilistic translation works better without the expertise of linguists.
McLemore said,
December 17, 2010 @ 10:02 am
Fred was a pill. Back in the early 90s when I was still often the only woman, or one of two, presenting at ANSI and DARPA meetings on lexicons, he was endearingly straightforward and willing to engage. Once he challenged me during a presentation in which I boasted about new pronunciation data for Mandarin [dialects], saying, "I find it hard to believe you and your staff know more than centuries of Chinese scholars." It was such a pleasure to blast him back and get his nod and smile. I'll miss hearing his gravelly voice on the phone looking for Mark Liberman….
Rod Johnson said,
December 17, 2010 @ 3:13 pm
Love the Hockett anecdote. I agree with Mark's assessment of its…evidential status, but it still rings true. I think both John's outrage and Jonathan's faux-outrage are a little misplaced. Those works Hockett published in later years had a heavy component of memoir, elder-statesman synoptic overview, reprints and why-in-my-day grousing. His most productive period was largely behind him by the sixties, and he may very well have been primarily focused on composing. With a lot of the old-school structuralists, as the center of gravity of the field moved Chomskyward, they found themselves somewhat adrift, and it wasn't uncommon to change focus or move toward more broadly humanistic pursuits. Ken Pike spent his last decades focusing more on poetry and 50000-foot essays on tagmemics than the kind of language description he was known for. These days he would probably have just started up a blog. As for Hockett's time at Rice, when you hire a scholar at age 70, it's often as much for teaching and mentoring (and name value) as original scholarship.
Rodger C said,
December 18, 2010 @ 1:16 pm
The first thing I ever read by Hockett was his notorious dismissive review of Chomsky's Syntactic Structures in Scientific American. He must have been increasingly lonely after that.
Bob Ladd said,
December 21, 2010 @ 7:01 am
I knew Hockett as well as any student did between about 1970 and 1980, and I can definitely confirm that he was doing roughly the work he was getting paid to do. I have always thought that the last decades of his career were largely misspent, but for reasons that are more tragic than outrageous. He never got over the Chomskyan revolution and was, as Rodger C says, increasingly lonely after that. The implications of his career for arguments about tenure are at best unclear.
BTW, I wasn't aware he composed operas, but he did do a lot of sonnets, and at least one march in the style of Sousa.
Rod Johnson said,
December 22, 2010 @ 10:39 am
He did write at least one opera that was publically performed, according to Wikipedia.
Rod Johnson said,
December 22, 2010 @ 11:05 am
OK, I've been looking for the "notorious dismissive review" and can't find it. Hockett did write a well-known article for Scientific American, "The Origin of Speech," in 1960, but at that point he was still enthusiastic about Chomsky. As late as his 1964 LSA Presidential address he was praising it as one of the greatest works in linguistics. At the same time, he was becoming more skeptical about the foundations of post-Bloomfieldian linguistics, which I think he considered Chomsky part of, so it's hard to say what it all added up to–Hockett was a notoriously changeable and complex person. It's hard to say when he definitively went over to the "opposition" but it probably wasn't until around Aspects (1965). So I would like to track down this review–Rodger, can you supply a citation? (Sorry for turning this into a thread about Hockett.)