User talk:Gekkoma

Machine Translation

Why don't we harness tools such as Google's language translation in the form of wikipedia bots to carry out bulk machine translations of articles? The imperfect articles can then be quickly 'polished' off by a human. Firstly, this is much faster and cheaper than individually translating tens of thousands of articles. Secondly, it rapidly allows for knowledge to be converted and transferred into other languages. Any feedback?

Cheaper? Whats cheaper than free? Wuhwuzdat (talk) 19:24, 19 June 2009 (UTC)[reply]

The opportunity cost of human time to translate. We are all ultimately volunteering - but in the time I spend 10 hours translating one document, I could quickly proofread 5. Thus my productivity on Wikipedia goes up in terms of translation. --Gekkoma (talk) 19:29, 19 June 2009 (UTC)[reply]

It might be against the terms of service to do automatic translations and reproduce the results on wikipedia. In google's terms and services see sections 5.3 and 5.5. Sifaka ^talk 00:51, 21 June 2009 (UTC)[reply]

There are plenty of other translation services out there... and considering it is for non-profit, a quick talk to google might be enough to OK the issue. Gekkoma (talk) 06:04, 21 June 2009 (UTC)[reply]

If someone could find a free translation service that gave reasonably good results. I could see a bot being created to do this. --Kraftlos (Talk | Contrib) 09:19, 22 June 2009 (UTC)[reply]

Why? Because they don't work. "I can proofread 5 documents"... Good! But can you understand what the machine was "trying to say"? Here's an example from Talk:Borki train disaster:

Google translate says "Make a project on the image and likeness of Christ Memorial Church of the Redeemer at Borki"
Yahoo translate says "To draw up draft on means and similarity of the memorial church of Christ rescuer at to draw up draft on means and similarity of the memorial church of Christ rescuer at station [Borki]"

Can you proofread such nonsense unless you can easily comprehend the original text? The original text said, in plain words, that a church building in another city was modelled after a church in Borki. Now you can. But not before that. NVO (talk) 07:52, 24 June 2009 (UTC)[reply]

p.s. Curious "image and likeness" and "means and similarity" are actually biblical "in our image, after our likeness..." (as in Genesis 1:26 KJV rendition) that became norm of vernacular language of the original. NVO (talk) 08:01, 24 June 2009 (UTC)[reply]

I would imagine that with this proposal, people who readily know both languages would still be involved. However, to save time, they proofread automatically-translated documents instead of doing the translating themselves. —harej (talk) 03:18, 26 June 2009 (UTC)[reply]

As has been pointed out above "free" online translations services have terms and conditions such as Google Terms of service, Yahoo Terms of service. Also the quality of translation is poor. These tools can be helpful to those with some familiarity with the languages, but only as a guide. Jezhotwells (talk) 21:29, 7 July 2009 (UTC)[reply]

I use them in two places, first, where I have a minimal language and am lacking some basic native vocabulary--and secondly, I rewrite some articles that have been submitted as machine translations from languages that i do not know in the slightest. I find I can almost always figure out from the present quality of free machine translation machine translation what is intended if I have any familiarity with the subject. What I want to ask, is does anyone here have experience with the actual commercial higher-quality programs rather than the crippled free ones? DGG (talk) 01:24, 12 July 2009 (UTC)[reply]