First, be careful not to confuse interpretation (spoken, real-time) with translation (text-based and delayed).
Most people commenting on machine translation take either of two extreme stances. My wife is a translator and we often have lively discussions on the subject. Some say it is the best thing since sliced bread because it will break down language barriers all over the world and help people communicate and put an end to wars and all that stuff. Others insist that it is impossible because it produces utter nonesense and is unreliable - "Look it made a mistake! This is so baaaaaaaad!".
I believe there are different levels of translation quality, each very well suited to different applications. Just like restaurants.
a) The ultimate where everything is perfect. This is where the text is handled by a professionnal and experienced translator with extensive knowledge of the field and reviewed by other professionnals of the field in the target language. You want this for contracts involving large sums of money, for written agreements between countries, for operating manuals of life-critical systems and the like. This could be assimilated to the best restaurants in the country, the five stars ones you find in posh travel guides. This is unfortunately what most people think about (and expect) when they hear the word "translation".
b) Normal, cost-efficient, human-made everyday translation. This is where the language is correct and there are no obvious mistakes for the casual reader. It often contains misinterpretations and missed jokes but the meaning usually gets through. This is what you get with translated news, instruction manuals of cheap, foreign made VCRs and internal specifications for international projects. You could liken this to the nice little family restaurant a few streets away. Nothing too fancy but it's better than what you would make yourself.
c) Machine translation where you can see the obvious mistakes and the language is often improper. But if it is not high end litterature, some of the meaning gets through. This is like most fast food. You can do better yourself but it's fast, cheap and convenient.
Machine translation opens a whole new market of "understanding" foreign texts. Although the result is almost never suitable for publication, it allows many people to have cheap and fast access to information they would never have used otherwise. And sometimes to spot pieces that are worth translating with a higher level of quality.
Machine translation will not bring perfect free translation to the masses and it will not put translators out of business. Just like McDonald's did not bring top gastronomy to your average shopping center and it did not put good restaurants out of business.
We should not expect perfect results form machine translation. It should instead be recognized as a very usefull addition to the tools at our disposal with it's own cost/efficiency ratio.
There is a nifty little utility that does just that and more...
It is called SETI Monitor and is listed on the main SETI@Home site under Related web sites - Add-ons for Windows clients.
You can get it at http://users.surfree.net.il/l.fainshil/
I am sure there are tons of other utilities that do the same thing but I like this one.
SETI Monitor displays a graph of the the spikes and gaussians found on the work unit currently being processed along with the estimated time left, estimated total time for the WU and other technical information taken from the state.sah file.
It also remembers all your past work units with all the above stats for each and can display a summary page of all the spikes and gaussians you found since it has been running.
I've been using SETI Monitor for a few months now and although the information it displays is interesting in the beginning, it slowly becomes almost as boring as the screensaver graphics. But I still like the feature where it keeps track of all the WUs that I have processed (and you can GREP through its list of old result files to do your own stats).
Slashdot Mirror
Most people commenting on machine translation take either of two extreme stances. My wife is a translator and we often have lively discussions on the subject. Some say it is the best thing since sliced bread because it will break down language barriers all over the world and help people communicate and put an end to wars and all that stuff. Others insist that it is impossible because it produces utter nonesense and is unreliable - "Look it made a mistake! This is so baaaaaaaad!".
I believe there are different levels of translation quality, each very well suited to different applications. Just like restaurants.
a) The ultimate where everything is perfect. This is where the text is handled by a professionnal and experienced translator with extensive knowledge of the field and reviewed by other professionnals of the field in the target language. You want this for contracts involving large sums of money, for written agreements between countries, for operating manuals of life-critical systems and the like. This could be assimilated to the best restaurants in the country, the five stars ones you find in posh travel guides. This is unfortunately what most people think about (and expect) when they hear the word "translation".
b) Normal, cost-efficient, human-made everyday translation. This is where the language is correct and there are no obvious mistakes for the casual reader. It often contains misinterpretations and missed jokes but the meaning usually gets through. This is what you get with translated news, instruction manuals of cheap, foreign made VCRs and internal specifications for international projects. You could liken this to the nice little family restaurant a few streets away. Nothing too fancy but it's better than what you would make yourself.
c) Machine translation where you can see the obvious mistakes and the language is often improper. But if it is not high end litterature, some of the meaning gets through. This is like most fast food. You can do better yourself but it's fast, cheap and convenient.
Machine translation opens a whole new market of "understanding" foreign texts. Although the result is almost never suitable for publication, it allows many people to have cheap and fast access to information they would never have used otherwise. And sometimes to spot pieces that are worth translating with a higher level of quality.
Machine translation will not bring perfect free translation to the masses and it will not put translators out of business. Just like McDonald's did not bring top gastronomy to your average shopping center and it did not put good restaurants out of business.
We should not expect perfect results form machine translation. It should instead be recognized as a very usefull addition to the tools at our disposal with it's own cost/efficiency ratio.
It is called SETI Monitor and is listed on the main SETI@Home site under Related web sites - Add-ons for Windows clients.
You can get it at http://users.surfree.net.il/l.fainshil/
I am sure there are tons of other utilities that do the same thing but I like this one.
SETI Monitor displays a graph of the the spikes and gaussians found on the work unit currently being processed along with the estimated time left, estimated total time for the WU and other technical information taken from the state.sah file.
It also remembers all your past work units with all the above stats for each and can display a summary page of all the spikes and gaussians you found since it has been running.
I've been using SETI Monitor for a few months now and although the information it displays is interesting in the beginning, it slowly becomes almost as boring as the screensaver graphics. But I still like the feature where it keeps track of all the WUs that I have processed (and you can GREP through its list of old result files to do your own stats).
Happy crunching.