Hi Samuel,
Thanks for the clarification. Good to know that the foundation is in the know.
Ravi and I have acted as interlocutors with the Google team for Tamil Wikipedia. We have exchanged several emails and have had one conference call with the Google team. During these communications, we have conveyed clear bullet-pointed requirements that are the bare minimum necessities to meet our guidelines and are very much doable. Of these, to be fair, they did address some of our issues, but not the most important ones.
The most important of the issues stem from the pillars of Wikipedia and we absolutely can't compromise on that. For Google, the required outcome is the number of words in Indian languages SEOed from their query logs. For the translators, it's the money that they'll get for each word translated. For Wikipedia, the basic necessity is readable and meaningful content added through a process that doesn't subvert the Wiki way.
Following is a summary: 1. The quality is abysmal. Too mechanical and ungrammatical more than 50% of the time. [To set the context for Samuel (who might mistake that it works like it does for European languages), the toolkit is not anywhere ready for Indian languages and doesn't do any translation as such, it's the translators who do that and it's unimaginable that a native speaker writes those words, not sentences.] 2. The process is hands-off, the translators don't even read the page that they've dumped. 3. The pages are broken with infinite erroneous redlinks and missing templates due to an easy-to-fix bug in the kit. 4. The basic premise of the team is 'something's better than nothing'. It's not. Having no article on a subject is better than having an unreadable text of 2000 words on that subject. 5. Their process requirement: you can pick subjects, give guidelines, but we can't guarantee anything. We don't carry any responsibility to improve the articles once dumped and we don't want you to mess with them. Of course, on the last point, they have come down. They agreed to have a look at talk page feedback and only one translator (of nearly 20-30) has responded so far. This is CLEARLY unacceptable and our editors have said it in as many words.
I also request the community here and the foundation folks to reflect on the policy issues: how can we let someone post articles of no acceptable level which they won't edit further? Tomorrow, if a vandal does the same, won't we block them? On top of this, they casually mentioned some sort of agreement or contract with the foundation, but decline to give any information regarding that. Either they don't get what Wikipedia is or they don't care about it.
On a positive note, we still have our channel open with them and we're going to propose that they approach universities or the Classical Tamil Institute in Chennai who undertake such projects employing retired Tamil professors and teachers. Also, carrying an obligation to fix issues before adding new articles. If they can't do that, we don't have any other option left.
- Sundar
"That language is an instrument of human reason, and not merely a medium for the expression of thought, is a truth generally admitted." - George Boole, quoted in Iverson's Turing Award Lecture
----- Original Message ----
From: Samuel Klein meta.sj@gmail.com To: wikimediaindia-l@lists.wikimedia.org Sent: Wed, April 21, 2010 11:41:16 PM Subject: [Wikimediaindia-l] Re: Philosophical view on Google translated articles
Hello,
My first post on this list, and a long one :-) The topic of
better
supporting small language Wikipedias is one that is close to my
heart.
The foundation doesn't have any particular policy on
third-party
translations or article-writing projects. As Achal says,
every
community is welcome to use translation tools or not as they see
fit;
and to work with outside translation groups or not as they see
fit.
Ravi's concerns are valid -- people interested in translation as
a
whole may want to discuss some of these issues on the foundation
and
translation mailing lists -- you will find that there are
many
multilingual editors who are interested in the good (and bad) uses
of
GTT and other translation tools.
== on the use of automatic
translations ==
Automatic translations can be useful as one arrow in the
quiver of a
community of editors. For instance, I find it helpful for
translated
pages to have an automatic category, and a large cleanup template
at
top, something like: "this page was automatically translated by
[TOOL]
from [permalink to revision of article in another
language].
It may need cleanup to meet [[STYLE GUIDE|community
standards]]."
In the case of Google and their Translation Toolkit, I
think it would
be good for Wikipedians to give them strong feedback about how
they
need to improve the tool for it to be more useful to
Wikipedians.
(and, if it is more of a nuisance than a help, the community
should be
clear that it is not helping.)
== On Google's toolkit
and translation work ==
Google has been fairly transparent about what
they are doing, and has
been in touch with the Foundation on a few occasions
to ask for advice
on how to make their tools more useful. I encourage
them to ask the
local communities directly for that advice... (however, they
have had
few direct responses from those language-communities. I
observed this
directly on swahili wikipedia - there were a few general
commnts about
the difficulties raised by GTT overwriting existing articles,
but few
specific feature requests / recommendations / requirements from
the
active swahili editors.)
You can start a page for feature requests
(and feature requirements)
for this sort of translation -- and tell the
Google translators (in
particular) that all translations /must/ adhere to a
certain style or
format, or must be less invasive when an article already
exists on the
topic. (noone will continue a project if they know that
its work is
going to be reverted or removed.)
From: Srikanth Ramakrishnan < href="mailto:rsrikanth05@gmail.com">rsrikanth05@gmail.com>
I agree with Shiju and Ramesh. I tried it out for Hindi. And the phrase 'A fully charged battery' got translated to what would mean a battery that got charged [the court charged]. It isn't all that accurate right now, but it may improve. While to a certain extent, it may seem like Google is catalising Localised content, you can clearly see that Google might be trying to gain Monopoly over Wikipedia as well.
I don't
think they have any interest in gaining monopoly over
Wikipedia. They
are not storing the translated articles, only
publishing them to
Wikipedia. While they are storing the "translation
memory" produced as
a result, they make that available under a free
license, for other
translators or tools to use.
Google has carried out similar projects
in Arabic and Swahili among
other languges; I helped with the recent
Swahili Wikipedia Challenge,
which was supported by GTT (for participants who
wanted to use the
toolkit to translate an article rather than writing one
from scratch)
-- but the resulting articles were rated based on their
usefulness, so
that poorly-translated articles did not rank
highly.
That was a largely community-driven translation effort, with a
contest
run and maintained by Swahili admins.
Cheers, SJ -- Samuel
_______________________________________________ Wikimediaindia-l
mailing list
href="mailto:Wikimediaindia-l@lists.wikimedia.org">Wikimediaindia-l@lists.wikimedia.org
href="https://lists.wikimedia.org/mailman/listinfo/wikimediaindia-l" target=_blank
https://lists.wikimedia.org/mailman/listinfo/wikimediaindia-l