As some of you know yourselves and others will hear from our localization tool vendors, progress is being made on the engineering front. We know--more or less-- how to go about it, we just need to slog away building the software edifices we need. However, on the other front, that of language processing, progress is slow--damn slow. You might even be getting impatient. But have you thought about why things are going so slow here?

The truth is that language processing is a very young, still immature field, in which we have large gaps of knowledge. Human communication is still not well understood. While surface phenomena such as morphology (word forms) and syntax (word order) have been reasonably well codified, and progress is being made with semantics (word meaning), there are still a great number of aspects of language, such as discourse and intention, which are still deep mysteries; this poses immense problems in teaching machines to handle language properly.

Language processing is an imperfect science--and is likely to remain so, at least in our lifetimes. For this reason, it requires substantial resources to indulge in, if only to be able to evaluate it properly. While a few very large companies and organizations have the resources for it, most of us don't. LISA is therefore an excellent vehicle for the inherent risks of these new technologies to be shared among its members. This happens informally, as its members exchange experiences anecdotally. But more important, it is one of the prime thrusts behind the SIGs, who are paving the way for the other members in the tough but much needed areas of evaluation and standards. Call it safety in numbers--for it is a matter of survival.