Saturday, February 21, 2004

Sally Thomason gets all huffy about speech engineers: I don't have any trouble understanding why many people who work on speech recognition and machine translation find much of linguistic theory unhelpful, because only some of what linguists know is likely to be useful for these purposes. But announcing the discovery of prosody (she was actually talking mainly about sentence intonation, not other prosodic features) seems a bit much.[Language Log] I wasn't at that particular talk, but I know about that general line of work, and I think Sally Thomason missed the point. it's not that speech engineers just discovered intonation, but rather that they are finally starting to figure out how to write algorithms that recognize intonation in useful ways. This is nontrivial, as the acoustic manifestations of intonation are complex and highly variable. I've found as a rule of thumb that a useful scientific result takes a couple of decades to make it into engineering, not only because of the obvious obstacles to the propagation of knowledge between fields (different communities, different technical vocabularies, different formal tools), but also because there are many tricky implementation details that are irrelevant in science but are utterly critical in engineering (such as the small matter of available computing power). As far as I recall, speech engineers have been thinking about intonation for the last twenty years. Probably not as deeply as a linguist, but still not ignoring it. But much that the human ear and auditory cortex can recognize accurately is exceedingly difficult for even the best current algorithms to recognize poorly. We just don't know how to separate the signal from inessential variation. 7:48:41 PM

It was interesting running across this newspaper article on the possible demise of the ski industry in Utah as a function of global warming.[tingilinde] The ski areas mentioned in the article as involved in lobbying efforts are the ones at lowest elevations in Utah. That makes sense, except that they are likely to be the ones that use most energy because of snowmaking. Snowmaking requires a lot of energy to pump water and compress air for the snow guns. Also, some of those areas (Park City, Deer Valley, Canyons) have the largest real estate developments, again not the best way to save energy, as many of those houses and condos are oversized and have huge windows for fashion's sake. It is well known that most ski areas focus on real estate anyway, so their pleadings are mighty suspect. On my backcountry outing in Utah last month, I noticed again the contrast between the overbuild, low-elevation Park City and Canyons, which were not located for the best snow conditions anyway, and higher-elevation, wilder areas like Solitude and Alta. Finally, the high-elevation areas there as well as most areas in Tahoe close in the Spring for lack of customers, not lack of snow. In summary, of course global warming is a matter of grave concern for everyone and not only for skiers, but the these ski areas are not the most convincing advocates for action given the contradictions between their words and their actions. 5:50:00 PM