I agree with what Belayet mentioned about the Google translated articles on Bengali Wikipedia. So far, we have not been contacted by Google directly. Rather, we have dealt with the paid contractors who were hired by Google. As Belayet has said, the paid translators do not follow up with the translations (except for a single case). Which in turn, causes a lot of problems for us to fix the articles.
Technically, we have not "banned" Google translated articles or contributors who use GTT. Rather what we have discouraged is the dump-and-run translators who just dump their malformed translation and never responds to our messages or makes a second edit to the article to fix it. In most of the cases we dealt with recently, we haven't even deleted the articles ... rather we moved them to user space, giving the user a chance to fix the article to readable Bengali.
So, basically, here are the issues:
1. We'd love to work with Google if they collaborate with us and take responsibility of producing readable content.
2. Which means, translations can't be dump-and-run jobs. Since the translator toolkit is still horrible in English-to-Bengali translation, the Google team or their translators need to do fix the pages to make them readable and grammatically correct.
3. We can use the current system ... i.e., translators are free to do all of the sandboxing/translation experiments in their user space. We are very picky about the content that goes into the article space and don't want half-done, incorrect language articles to go there. So, after a translation article has been approved by the community, it can be moved to the main space.
But to do any of the above, the Google team needs to contact us. We don't know who is or who isn't working for Google. Almost all the vendors/contractors we have dealt so far used thorwaway accounts that are used only once, and never again.
Bottom line: we are happy to work with Google, but only if Google does not bypass the existing Bengali wikipedian community.
Thanks,
Ragib
User:Ragib on bn and en
-- Ragib Hasan, Ph.D NSF Computing Innovation Fellow and Assistant Research Scientist
Dept of Computer Science Johns Hopkins University 3400 N Charles Street Baltimore, MD 21218
Website: http://www.ragibhasan.com
On Thu, Dec 2, 2010 at 9:19 PM, Belayet Hossain bellayet@gmail.com wrote:
Ravi, That's a nice process to deal with Google translation project. In Bengali Wikipedia, if the translation is not in acceptable quality community also shift content to the user namespace of that translator and ask them to improve it. But there are very few examples that translator rewrite or retouch the article to improve it. Translators are not coming back to take care of their article. So a lot of untouched bad translated articles are in the user namespace at Bengali Wikipedia.
And from my experiences in Bengali Wikipedia the translators are not consistent with their translations. If you rated someone for the first time for his first translation, it is not be sure the second translation will be same quality or better than the first one. So community have to re-rated him every time he post an article.
Since the translators are not regular at Wikipedia and they are not responsive. There is no other contact point available for the community to communicate with them. We can create a translation coordination page at local Wikipedia, but there is no way to inform the existing or new translators to follow the page.
I am very much interested to know, how Tamil community communicating with the translators at Google?
Belayet
On 3 December 2010 07:12, Shiju Alex shijualexonline@gmail.com wrote:
Congrats to Tamil community for trying to bring out a process for Google's Translation project.
I really wonder how other language communities are handling this. Apart from Tamil, Google Translation project is going on at least in Hindi, Kannada, and Telugu. It is banned in Bengali wiki.
I could see many articles are loaded to wikis each day. And for many of them the only contributor is the Google employee who translated it.
Shiju
On Thu, Dec 2, 2010 at 7:55 PM, Arjuna Rao Chavala arjunaraoc@googlemail.com wrote:
Hi,
Thanks a lot for the update.
I think the updated process is similar with Open source community philosophy when Commercial companies (like IBM, Sun etc) contribute source code. Tamil Wiki has that kind of rigor in quality checking and is able to do a good job. Other Wikipedias may not be in a position to engage in a similar way, due to policies and/or level of active wikipedians. One more comment below.
On Thu, Dec 2, 2010 at 7:19 PM, Ravishankar ravidreams@gmail.com wrote:
Hi,
Some updates on the Google translation project in Tamil Wikipedia. --snip
We did a quality review of these articles and found that only around 50% of them has an acceptable minimum quality regarding translation ( We just rated the style of the language and accuracy in translation. We did not do a full review on the merit of the article). --snip--
Can you elaborate more on the style? How did you measure the accuracy of translation? It may be desirable to adopt the English articles for the target wikipedia, than verbatim translation. How much effort was spent to arrive at the above conclusions? # of articles, # of reviewers, time frame etc would help.
Thanks Arjun
Wikimediaindia-l mailing list Wikimediaindia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikimediaindia-l
Wikimediaindia-l mailing list Wikimediaindia-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikimediaindia-l
-- Belayet Hossain http://www.facebook.com/bellayet http://twitter.com/bellayet http://bellayet.wordpress.com (Bangla) Knowledge is universal ...so share it.
Hillel____ If I am not for myself, who will be for me? If I am only for myself, what am I? If not now, when?
Wikimedia-BD mailing list Wikimedia-BD@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikimedia-bd