Thursday, September 25, 2014

N'Ko on the web: Review of experience with ebola FAQ

Here's a quick recap of the demonstration of (experiment with) posting the N'Ko text of a WHO FAQ on ebola. Shortly after I posted it, I wrote "There would seem to be no reason not to use the internet for dissemination of webpages and mobile content about ebola in N'Ko, ..." Then Charles Riley of Yale University and Athinkra LLC pointed out some problems. And I found others.

I stand by that estimation, but after some tedious work on formatting and font coding in HTML, do so with greater emphasis on the caveats: "... although there would need to be attention to testing of commonly used systems and of ways to feed or facilitate loading of fonts that include N'Ko."

A summary of findings and lessons learned follows, but first a quick recap of what I did. My first attempt was to copy-past N'Ko text from a PDF of the FAQ on ebola, which as previously mentioned, did not work (RTL text was switched to LTR; character order was mixed, and combining diacritics sometimes not combined). Working from the Word document original, in contrast, was deceptively easy since I was working in Firefox (FF) ver. 32 - copy-paste into the "Compose" screen on Blogger, fix some bullets and do some minor formatting and voilà!

Problems were identified in Microsoft Internet Explorer 8 and Chrome 37 (mostly white boxes but some strings of text), and FF (some alignment of combining diacritics used to mark tones). Varying font commands inherited from the Word source document caused the irregular display of characters, and were fixed (an alternative to try later would be an unformatted text paste). 

However, while this work facilitated a correct display of diacritics on FF 32.02 and MSIE 8 on a computer running Windows XP, the diacritics were all off the mark in FF 32.03 and MSIE 11 on a Windows 7 system (the latter did not have the DejaVu fonts installed, which might be the problem there).

Lessons learned so far:
  • align="left" - Full justification of N'Ko text may space the combining diacritics as if they were characters, leading to misalignment.
  • Font choice/availability - N'Ko text display seems especially sensitive to font commands. Do not have an answer to why formatted text from a Word document pasted into Blogger looked fine in FF but not in MSIE 8 or Chrome. Installation of a font with N'Ko support may help.
  • Browsers may not be the main issue - Despite initially encountering display issues in MSIE 8 and Chrome more than FF, those could be corrected in the source code. Also, the fact that the most current FF (32.03) and MSIE (11) misplace all diacritics on a computer without a particular font (DejaVu) points back to the font issue. 
  • But browsers are not not an issue - The N'Ko script normally shows liaisons between characters within words (somewhat like Arabic), but while these show on FF, they did not on MSIE
  • No bold or italics - Bold N'Ko is apparently not supported by the DejaVu font. Italics are, but while those did display on FF 32, they would not on MSIE 8.
Next steps

This has been a learning experience, but it should be pointed out that with sites like Kanjamadi, N'Ko on the internet is a reality and a potential to be pursued, even as there are issues. (Kanjamadi displays impressively on the same FF32 on XP that did great with the WHO ebola FAQ, but cannot load in MSIE 8 and has the same diacritic problems on MSIE 11 on the Windows 7 machine mentioned above).

Before posting more in N'Ko, it would be helpful to have more feedback on display issues for this script in function of available fonts, browsers, and operating systems.

Next steps should have as their goal a simple how-to for organizations wanting to display text in N'Ko on the web - whether for ebola education or any other useful purpose. Similar localization guides could be developed for other West African languages as well.


Howard Gutowitz said...

As you probably know, our approach in the app world is to do both: N'ko unicode + image of the text (for email, posting to Twitter, Facebook, etc). Those that have full support of N'ko can enjoy it that way, for those that don't, there is a fall back. In general, the way forward is often to supply backwards compatibility. -- /hag/ for Eatoni.

Don said...

Thanks Howard. Yes, for dissemination in complex scripts, fully agree that an accompanying image file alternative/fall-back makes sense. Have also used Latin 1 text alternatives for complex Latin.

For revision and re-use (2 of the "4R's" I plan to post on soon), getting text to work as widely as possible seems is of course important. The case of N'Ko is interesting given its recent encoding, and its relatively small and non-affluent, but very active potential user community. Technically there's no reason it couldn't be supported in text as seamlessly as say Arabic, but the economics of N'Ko don't favor that level of attention. So in the meantime, I'm thinking it's useful to see where the "boundaries" of support are for this script.

BTW, am interested to note Eatoni's coverage of African languages. Have you done anything with N'Ko? Noting Bambara/Jula, do you have any products supporting its extended Latin orthography?