[language, technology] small delight
Dec. 26th, 2023 09:05 pmI have for many years now used the Swype keyboard for text input on my auxiliary internet device. Languages I have installed are English (UK), French, German, Irish, and Turkish.
One of these languages is not like the other, not merely in that it's not Indo-European but in specific that it's an agglutinative language.
The way this is handled in predictive text (1) makes a lot of sense and (2) is brilliant.
For four of these languages, if a word isn't in the dictionary (whether because it came pre-entered or because I saved it myself later), predictive text won't ever suggest it. "Stork" can be in the dictionary, but if it's only there in the singular I can swipe that shape + "s" and I'll never get "storks" suggested.
By way of contrast, have a word I used as demonstration to A the other night: duşmanlarımdan. "Duşman" (enemy) is in the pre-loaded dictionary; even duşman+lar (enemy + plural: enemies) might be. But duşman+lar+ım (enemy + plural + first person singular possessive: my enemies) isn't, and duşman+lar+ım+dan (enemy + plural + first person singular possessive + ablative: from my enemies) definitely isn't. The fundamental nature of the language, though, is that you do just Build Words like this, so rather than try (and fail) to preload All Possible Words or even a reasonable subset of them, the Turkish keyboard layout has instead been written to make predictions based on what set of base word plus standard morphemes you've swiped over. It's lovely.
(The target sentence in question is "duşmanlarımdan kaçabilirdin", "you could have escaped from my enemies". "kaçabilirdin" breaks down to kaç-a-bilir-din: "kaçmak" is to flee or escape, and gives "kaç"; "a" is a buffer vowel, which follows vowel harmony; "bilir" is to be able; "din" is the second person singular past tense indicator.)