Re: New languages



With regards to adding new languages, FlickType is a bit unique due the nature of how much inaccuracy it is designed and expected to tolerate. Simply grabbing an existing dictionary is not enough, because there are often valid words that are very uncommon, but have letters similar to words that are much more common. One such example for English is the word "fir", spelled F I R and referring to the tree, which is extremely close to the letters of the word "for", F O R. It would be very frustrating for users if there's even a small chance they get "fir" when they try to type "for", because that small chance would be amplified by just how common the word "for" is. To address this, we need to have a good understanding of how common or uncommon words are in each language we support. For a language like English, we have a pretty good idea ourselves of what words are common and what are not, although even with English there are some slang terms and jargon that are only common for some people and some areas, but not others. Language modules of other keyboards are not publicly available for us to use, and language modules that are available are usually from public sources like Wikipedia or news articles, which do not capture the way people use the language when they type on their device. For example, the term "GOP" that refers to the Republican Party in the US would far surpass nearby words such as "top" and "hop", while "lol" would barely even exist. So what we need is not as readily available, or even as objective as you might expect, even though there are some workarounds, each with varying levels of required effort.

All in all, I'd say that adding new languages to FlickType compared to other keyboards is not qualitatively different, but it's quantitatively harder since we need to have more accurate word frequency data than the average keyboard.

- Kosta

Join to automatically receive all group messages.