Invisibility and the Marcus parser

Philip Greenspun's Homepage : Philip Greenspun's Homepage Discussion Forums : NE43 Memory Project : One Thread
Notify me of new responses

I don't know when I first moved into the AI Lab, 1982 or 1983 maybe, thanks to UROP (the Undergraduate Research Opportunities Program). I don't remember any outright rudeness, as Peter Szolovits describes; rather, the pervasive ignoring, but I assumed both that I deserved to be ignored and that I wasn't singled out to be ignored: that it was a general ignoring. I moved into an office with Monica Strauss and Chris Hanson. I don't remember Chris being there much; I imagined that he spent most of his time in a machine room. Monica was something of a mystery, well-connected and an outsider at the same time. I assumed those two characteristics of hers were not unconnected; I've used the combination successfully myself, being one to be the other. Monica was an outsider because she was observing the lab, a sociologist studying the practice of our trendy science, comparing us to the Japanese and tracking AI Lab political intricacies that she hinted at to me conspiratorially from time to time. (I, on the other hand, was always highly oblivious to AI Lab politics, the way I was oblivious to compiler and operating system details.) I never had the impression that she was studying me; I was too new and too insignificant to figure in the tale; I was more in the line of audience. Monica was on networky first-name basis with all the Big Names. I found that surprising (since I couldn't imagine being that way myself), though I suppose it is flattering to the powers that be to have people studying them, writing about them.

I was at the AI lab thanks to Bob Berwick, because natural language processing was one of my favorite things in the undergrad AI course I had just taken, a course I loved. I loved the final, with the problem about how to distinguish one extraterrestrial from another, from an overhead image, using your machine vision skills to disambiguate the round heads from the cone heads based on Martian skin albedo. I wasn't actually getting the impression that AI was terribly useful, but it was really quite enjoyable. Anyway, I liked natural language processing and so gave up my physics UROP over at the tokamak where I had dutifully worked in the machine shop to build an alarm circuit into a metal plate, with a nice big red light that could flash on and off. The plate was mounted on metal frames alongside many other similar plates, a whole wall of circuits, switches, knobs and indicator lights; a childhood dream come true for me. My favorite part of "Charlie and the Chocolate Factory" was the lovely elevator with buttons everywhere. I was a declared physics major at the time, but my physics advisor, a thin nervous older total nerd professor, never looked me in the eye and was generally very discouraging; I think this was probably because I was a woman, but regardless of the reason, I was put off. It was the same feeling I got when I received my Rush week anti-tour of Bexley dorm: don't live here, you're not wanted.

But somehow I was not to be put off at the AI Lab; I solved this by being mostly invisible, and by enjoying myself, and by getting lost in just watching some of the amazing flaming -- in those days it was the philosophical rhetoric of Agre and Chapman and the like. Remember that way they and others had of talking? I wonder where it came from -- who started each verbal idiosyncracy -- how it was picked up on -- what did it signify? How to describe that style? Help me out here. Big philosophy words; a habit of filling frequent long otherwise-pauses with a drawn-out "umm", to keep the floor, I suppose. I always found these conversations over my head or outside my realm and anyway I was very shy and found it a big effort to walk into a playroom that had people in it.

Bob Berwick was very nice, and set me to investigating the applicability of the Marcus parser to the french language; as I recall, there was a theory about being able to parse any sentence with a very limited look-ahead, and French posed a possible challenge to this theory because of the clitics, or many small words including diret and indirect objects, reflexivity indicators and adverbs, that often show up in front of the verb. I don't remember what my conclusions were, but I suspect they were ambivalent; I have a deep tendency toward ambivalence, not wanting either side to feel that it has lost out. But thinking now about french clitics, it seems to me the finite register theory ought to hold for French despite the clitics, though you might have to throw out the autonomy of sentences: expand the notion of syntactic parsing to allow ties to the previous sentence. Maybe. Parsing is such a puzzle pleasure. I always loved language; I loved neat pages of handwriting in a foreign language, especially if the alphabet was cyrillic, like a code, and the alien sound and feel of the sentences in my mouth; the challenge of grasping and reproducing something quintessentially of that language -- a turn of phrase, a colloquialism, a native shorthand pronunciation. Latin was fun for the decoding -- first find the verb by the endings, and double underline it; a single underline for the subject; parentheses for the direct object, and so on, and then the meanings of the words, and fishing out the subtle emphases from the word order and even from the meter. A puzzle.

So it was natural to go and talk to Bob, and to work on the Marcus parser in french, and to have ambivalent results.

It was because of this project that I discovered I was spied on. Gregor Kiczales said to me one time that I was a prolific programmer or something like that, which I took as being extremely complimentary. I asked him why he would say that, and he referred to all the lisp code I had written. Now I probably had written some lisp code at that point, but I certainly hadn't told people about it except maybe Bob, and I therefore suspected that he had looked in my directories and found all the Marcus parser code that I in fact had not written.

I have sometimes wondered since then if anyone ever snooped on my email correspondence, which over the years was full of mundane juiciness -- gossip, fights, jealousies, beggings, outrage, anger, mortification, daydreams, stories, and the occasional technical communication. I must say I received much more eloquent email than I ever sent, most notably from Devon McCullough and Bruce Donald, who are both wonderful emailers.

-- Barbara Bryant, January 4, 1998

Answers

Wow... I barely remember being in that office. It was my first office when I came to the lab in 1982.

It's true I wasn't around much -- I spent all my time up ninth floor, trying to find a Lisp machine not in use by RG or RMS, or some other much-more-important hacker than me. Then a bunch of us went off to HP for the summer, so I _really_ wasn't around. And when we came back, I had a new office, just outside the very same playroom where I got to listen to all the very same flaming that I also didn't take part in.

Those were the good old days. Now, we have new flamers who flame about dull things like operating systems.

-- Chris Hanson, January 4, 1998