Fitaly is optimized for Indoeuropean languages

Fitaly was initially optimized for the Brown corpus for the English language. Later on, we had to address the question of how well Fitaly would do for other European languages. This question was actually raised by one of our German beta-testers early on. So we ran our tests for German on our keyboard simulator and the results were quite surprising:

Keyboard	Travel	HM
Qwerty	3.18	0.88
Fitaly	2.17	0.36
German	2.02	0.32

In the above, Travel is the average travel between keys (1 between adjacent keys) and HM is the number of hand moves per key (many hand moves are required by things such as punctuation). German refers to a keyboard that would be completely optimized for German.

What these show is that there is a huge difference between Qwerty and either Fitaly or German: a factor of 3 for hand moves - the really important parameter. On the other hand, the difference between Fitaly and German is similar to the difference you get by changing the subject: similar to the difference between administrative language and technical language.

So our conclusion was that the difference was not sufficient to warrant all the drawbacks attached to having several layout. I had already found out this for French and Italian. This must be a reflection of the common Indoeuropean origin of these languages.

Incidentally, we ran the above German tests on 2 MB of the writings of Franz Kafka. We then ran the tests on 300 KB of legal letters and the results are within 1-2% when you remove numbers, within 5% with numbers kept.

Keyboard	Travel	HM
Qwerty	3.16	0.87
Fitaly	2.17	0.34
German	2.03	0.31

Again, the conclusion is that although we could improve it slightly by a specialization to German, the marginal benefit is not worth supporting several layouts and having users learn different layouts when they converse in different languages.

French and Italian are even closer to English - in terms of letter distribution - so that they warrant even less a different layout.

Jean Ichbiah