For the past three days our company has been in southern california showing off our robot to the Army and a whole lot of MDs, PhDs, and in several cases MD PhDs. Some stories:
We anticipated having to pay some fees for taking our 7 huge boxes of robot parts on the plane. But a curbside check-in guy walked up to us quickly and said in the most suspicious way possible "if you work with me, I'll work with you." As he confidently put our boxes onto a cart, we all wondered whether he had any real power to decide what was checked baggage and what would be considered excess baggage and slapped with fees. He also assured us that taking 2-3 bags each as carry-on wouldn't be a problem. In the end all we could do was hope for the best, as we 'worked with' him to the tune of $20 (after handing him $15, he assured us "that's not gonna do it"). Fortunately, we got through unscathed, so it was a good deal after all.
We ate dinner in the hotel restaurant the first night, and when the waitress came up she said "wow, this is going to be a big bill!" She wasn't wrong, but I've never heard a restaurant employee comment on that before. I suppose parties of 4 are not very common. When she brought us the check she made another comment, this time "hey, I guess you didn't spend too much after all." I for one appreciate her financial candor.
One of the exhibitors is a pair of scientists from Kodak, presenting a stereoscopic image system. Apparently the benefit of it is that one doesn't have to wear goggles or special headgear to view it. Of course, one does have to put one's head in front of a giant plastic case that contains two LCD monitors, with the whole setup inside a black-walled enclosure to keep out reflections. But hey, no goggles! One of my coworkers got a 'special' demonstration of the system that featured stereoscopic images of nude women, which he described as 'kind of disgusting.' Science, huh?
There are a couple of other interesting/bizarre exhibits that I would not be doing justice if I did not wait until I can upload and include photographs of them.
Today another attendee who spoke at length with us told me of a project she had worked on which sounds to me like the hardest natural language processing task I've ever heard of, to an extent that it is nearly unfathomable. It was a telephone intepreter system, in which one user would speak into the phone in english, a speech recognition system would take the speech to text, a machine translation system would translate it into japanese text, and then a text-to-speech system would speak the japanese text to the other user. In this way, each user would be carrying on a conversation in their native language. So this includes two NLP tasks that are incredibly challenging (the STT and machine translation), one that is not too hard to do, but very hard to do well (TTS), plus it's using English and Japanese, two very difficult languages to translate between in about 100 different ways. For one, there is the problem of information being left out. The complete Japanese sentence "ikimashita." literally means "went," and the machine translator would somehow have to know who went to include it in the English version. Meanwhile, the English would contain no information about the relationship between the two speakers, except in very subtle ways. This might not be a barrier to understanding for the Japanese speaker, but it might leave them a bit shocked. The difficulty of these issues alone is insane, and as this woman told me, basically requires the software to have a knowledge base including everything about both users and their respective cultures. But it's cool that it was even attempted nonetheless.
In fact one cool aspect of this conference is seeing that women are breaking into science and heading up some of these projects, and that they have more degrees than the temperature in new york right now. When I last saw my stepbrother the philosophy professor, he asserted that philosophy, not computer or any other science, is the most bereft of women at high academic levels, but that no one seems to care!



Comments (4)
Interesting post... do you know any more about the Japanese/English real-time translation project, like, where it is based?
Cheers,
Pat
fieldmethods.net (returning from the dead, soon)
January 16, 2004 5:26 AM
having both an MD and PhD? what a waste of time! (not that the value of my PhD program isnt dubious enough)
January 17, 2004 11:13 AM
I thought that project was something IBM had been doing/working on for a while now.
January 17, 2004 5:21 PM
The project in question was based at ATR International in Kyoto. Judging by their website they appear to still be working on it and many associated technologies, but it's hard to say how far they've gotten.
I wasn't aware of it, but yes IBM is working on the same idea. Their system is called MASTOR and is intended to accomodate many languages and run on Pocket PCs for face to face communication rather than telephone. One item said that the approach was very 'context-dependent,' which seems to indicate that they at least acknowledge the problems of included and omitted information that I mentioned. It will be interesting to see if they have a solution for it, other than making their software omniscient.
January 18, 2004 3:38 PM