[maemo-developers] [maemo-developers] Asian language support?

[From: Aaron Kavlie]

Mr. Chen,

First of all, great to get a response here from someone who  
understands these issues well.  I appreciate it.

On May 31, 2005, at 7:43 PM, <xun.chen at nokia.com>  
<xun.chen at nokia.com> wrote:


> BTW, do you know any good quality Japanese fonts available from OSS  
> or commercial vendors? Good quality means not only the good  
> aesthetic looking, but also the less memory, especially RAM  
> consumption, and quicker rendering.
>

Been a while since I was on Linux (using OS X now), so had to go back  
and do a  little research: the Kochi Gothic (http:// 
packages.debian.org/stable/x11/ttf-kochi-gothic) and Kochi Mincho  
(http://packages.debian.org/stable/x11/ttf-kochi-mincho) fonts are  
what's available for truetype fonts on Linux.  I think they look  
fairly good, have no idea how they fare on memory & rendering  
compared to the commercial alternatives that ship with Windows & OS X.


> You are right, it is tricky to input Japanese or some thing like  
> that. We 'planned' to put Chinese input methods (VKB/HWR) in next  
> product. Like Japanese, Chinese input needs, so called, 'pre-edit'  
> area to convert Latin letters to kanji, also some candidate list is  
> needed. They become more difficult, if there is no hardware keyboard.
>

But proper handwriting recognition becomes an alternative to this  
latin-to-kanji conversion; you just write out kana/kanji directly.   
This is the ideal solution when not using a hardware keyboard I think.


> We knew the advantage of Decuma that has been bought by Zi Corp  
> last year. The key of the HWR is relying on HWR engine, the  
> algorithm to convert handwrite trace to text encoding.  
> Unfortunately, there is no such engine is available from OSS yet,  
> NOT even with non-product quality.
>

Actually, im-ja (http://im-ja.sourceforge.net/) includes handwriting  
recognition based on KanjiPad (http://fishsoup.net/software/ 
kanjipad/), which in turn is based on Todd Rudick's kanji recognition  
algorithms.  I have also seem his work implemented in PAdict (http:// 
padict.sourceforge.net/) for PalmOS.  I have not used it enough to  
attest to its quality, but as the variety of implementations attest,  
it is probably the best (perhaps only) available OSS code for  
Japanese kanji recognition, and would probably work fairly well if  
someone created a GUI for it.  As a starting point, at the very  
least.  Kana still needs to be added, but that should be far easier  
than kanji.

And then there's the previously mentioned Sharp Zaurus stuff, but I'm  
not as familiar with that.

-Aaron