Barrier to entry (was Re: Rosetta)

[Dwayne Bailey]

Been off sick - still a bit ill...

On Sun, 2004-10-03 at 00:12, Mark Shuttleworth wrote:
> Dwayne Bailey wrote:
> 
> >- The portal lowers the barrier to entry which I think is good but how
> >low should be make the barrier?  You have to be able to code to
> >contribute to the Linux kernel.  Otherwise you're shouted our of the
> >arena.  Maybe we shouldn't be so harsh but still how do we ensure that
> >people who contribute can actually translate?  A source project will not
> >be able to evaluate the quality of supplied translation.  I know the
> >Welsh team had a two stage process - but would like to hear other ideas.
> >  
> >
> One thing I'd like to see is an automated assessment of the difficulty 
> of translations. A translation gets "difficulty points" for each of the 
> following:
> 
>   - it is a plural

This is very confusing to new translators and a source of many problems
- removing them would help greatly.

>   - there are complicated plural forms in that language

There are not many languages with this issue so this might not be worth
the effort probably simpler to just leave plurals to better translators.

>   - maybe a language-modifier, some languages are harder than others

But you'll only be translating your language so if its hard its hard not
much you can do about that.  If you are translating two languages I
would be very concerned about the quality of the translations - its
inescapable you will be mixing language.

>   - the string includes embedded substitution codes that are programming 
> language artifacts

I assume from the last you means: variables, xml, html, etc.  These
after plurals are also the best source of incorrect translations.  Some
that I observe that cause problems:

- Variables - Some people don't realise that you change the order for
your language. Some people translate the variable names
- HTML/XML - tags, entities, attributes get translated.  URLs in <a>
tags get translated.
- File paths - some need to be translated when they are an example
(/home/bob/file.png) while some cannot be translated (/var/log/messages)
- Environment variables - TEXT_EDITOR

People do learn to recognise them over time and more technical
translators don't have as big a problem.

> ... we could explicitly present new users with "easier" strings, and 
> have a way to test and promote them to a more masterful state where they 
> could see any translation.

Short strings first
-------------------
I have thought of ideas of presenting users with short strings first. 
This has the added advantage that it would result in a visual appearance
of complete translation quite quickly because most long strings are
related to error messages and other help text not labels, titles,
dialogue boxes and tooltips which are all usually quite short.

The problem of flow
-------------------
The problem of grading translations is that it is usually quite
important to know the flow of the translations ie what came before and
what came after to understand how you should translate the current
item.  If we drop a plural string or do short strings first then we lose
that flow.  I would suggest then that we only present contiguous blocks
of easy strings so that there is a flow.

Beginners only files or hand-holding
------------------------------------
Another grading might be that you simply are classified as a beginner
with a more advanced translator tasked with reviewing your work more
thoroughly and providing useful feedback until you complete say 100
translations.  Or a group of files available only to beginners.  

Classifying files as core of free for all
-----------------------------------------
Another strategy would be to classify PO files in the translation
landscape.  So you classify core items as needing to be translated by
those with the most experience.  Beginner translators are then able to
work on anything on the fringe but not on the core until they have
enough experience under their belt.  How to define that experience I do
not now.  Probably best done by the people themselves.  So as an
example: core parts of Gnome are reserved for experienced team members
while items such as Gnome games are free for anyone to translate.



-- 
Kind regards
Dwayne Bailey

dwayne at translate.org.za   
+27-12-343 0389 (home/work)  +27-83-443 7114 (cell)

"It would be a profound irony if an earnest attempt to bridge
the digital divide unravelled because of prohibitive software
license costs.  Even with educational discounts and so forth,
the proprietary model does not offer the unfettered choice to
participate in the development or modification of the very
technology that can only increasingly become an intimate part
of any developing society as it ventures into a digital
future." - Dr Sibusiso Sibisi, President CSIR, South Africa

Translate.org.za - Opensource software for all South Africans
A project of the Zuza Software Foundation