Teutonizing the Newton

Udo Flohr

When developing software for today's global market, developers must contend with various factors, including adding support for non-English characters, translating text strings, and supporting varying time and date formats. But in working on the German version of the Newton PDA (personal digital assistant), which was released in mid-December, Apple's Newton localization team, headed by Cindy Roberts, found that handwriting-recognition systems add a new set of factors to consider in localizing a product. Apple's programmers had to contend with major differences between American and German handwriting and the languages themselves.

When you use the Newton's dictionary-based recognition system, you get a mixture of surprisingly good--and confounding--results. The dictionary-based approach, which considers only w ords that you have stored or added to the system dictionary, works well for languages like English. But this system doesn't account for inflections, which are characters that change a root word's tense, case, or number (e.g., the American version of the Newton needs separate dictionary entries for house and houses). Consider the following:

the green house/the green houses

das grune Haus/die grunen Hauser

la maison verte/les maisons vertes

The two phrases in three languages have equivalent meanings, but the German and French versions show more changes to the words. Newton would cover the English version with four dictionary entries, but it would need six entries for each of the other versions. Verbs have many more inflectional forms in German and French, so the system wouldn't require you to make as many corrections if the German version provided support for inflection rules.

But what the Newton localization team found even more haunting were those famous long German words. In Eng lish, compound nouns are written as separate words, but in German, they are written as one word. House door key would require three dictionary entries in English, but the German language needs a separate entry for Hausturschlussel, as well as for other permutations, including Haus, Tur, Schlussel, Haustur, Turschlussel, and all their plural forms.

German-speaking Newton team member Peri Altan refined these word lists and came up with a dictionary of 17,000 words, almost twice as many as the U.S. version (the Japanese Newton, which was due to ship in January, needs 8 MB of ROM--twice the ROM capacity as that of the U.S. version). Other changes that the Newton localization team had to add to the German version are listed in the text box "From English to German."

Localizing systems that rely heavily on handwriting recognition requires much more than the simple translation of text strings, but the same is true for systems that rely on speech recognition. For developers who are worrying about the wor k involved in localizing their Newton applications, Apple's Roberts says that many German and other international users will be happy with a U.S. version of an otherwise satisfactory program, provided the program does not force constraints (e.g., an inflexible phone number, ZIP code, time, date, and other formats) on the user.