14 December 2009
In the continuing saga of our information overlords, they’ve come out with Google Translate. As a former Latin teacher, I mostly love and partly dislike this system:
+ The on-the-fly translation is pretty sweet. In particular I love seeing how it recalibrates its concept of whole phrases as it gets new input — something I would have liked to have shown my students as a good practice.
+ It supports a bunch of languages and lets you choose any pair of them for initial & target (including some helpful options for non-Latin scripts and Romanization).
– Latin is not among the languages it supports, which limits my ability to probe it.
-/+ Using a language I know less well but can hack at lower levels (Spanish), I can see there are definite (and unsurprising) weaknesses, especially as sentences get longer (and presumably as grammar gets more complex, although once that happens my ability to translate the Spanish is also hampered). So minus, it doesn’t work as well, but plus, it still won’t be supplanting anyone’s language-homework-doing any time soon ;).
-/+ It uses statistical patterns derived from really big corpora (as we might expect of Google), not computational rules. On the one hand, my inner linguistics nerd is sorta sad. On the other hand, it’s awesomely googly (and more pragmatic/scalable, I’m sure).
However! The library angle I was getting at here is that you can search web sites in other languages. Enter terms in the language you know, and it’ll translate and search. Looks like it will only search one language at a time, and I don’t know how it deals with ambiguous terms, and I’m sure the quality degrades with phrase searches, but this does increase our ability to find all relevant information on a query, and I’m sure the tools will improve with time.