Putting Linguistics Back Into Computational Linguistics

Speaker:
Martin Kay
Abstract:
The belief has recently become widespread that  the properties of language needed to process it for useful  purposes will emerge if sufficiently large quantities of raw  text and speech are analyzed automatically using sufficiently  sophisticated techniques. The kind of understanding that a   linguist attempts to achieve by examining individual  specimens at close range has little value, at least for  practical purposes. But, if information can be caused to  emerge from the raw data only if it is in there in the first  place, and it has long been known that this is not the case. A   language is a code, that is, a system of arbitrary relations  between symbols and things in worlds, real and imaginary.  No time or effort invested in examining the symbols will  reveal these relations to one who does not know the code. If  this is true, then we must ask why statistically based  machine translation, for example, has come as far as it has,  and how much further it can expect to go.
Length:
01:07:17
Date:
26/04/2013
views: 924

Images:
Preview of img-001.jpg
Image img-001.jpg
Preview of img-002.jpg
Image img-002.jpg
Preview of img-003.jpg
Image img-003.jpg
Attachments: (video, slides, etc.)
62M
330 downloads
104M
323 downloads
452M
316 downloads