This must indeed be the ultimate goal. However until we get to that stage the best approach, in my view is to have separate pages for translated text.
http://sites.google.com/site/bwlangdf/home/francais Consider this. This was arrived at by going to Google Translate and editing. The sentence that was wrong was :- "Il fut fondu par une groupe d'amis en 2010 pour trois raisons" Google rendered this to Elle a été fondu. Well the Past Historic is clearly better than the Perfect in this context. More serious though is Le magazin and "Elle". This sentence proves that editing is still needed. Another 2 examples this time Arabic - English https://sites.google.com/site/aitranslationproject/deepknowled If I look for the Stefan Boltzmann law and ask questions (jeopardy style) I get the correct law. Google for مضروبا في اربعة اضعاف درجة حرارة سطحه اى gives me " Multiplied by four times the temperature of any surface" حرارة which means "power" in this context is not translated. It is a moot point whether it is possible to produce really good translations without a level of understanding. This is a poor example as the first criterion is to translate all the words. Microsoft translates this the same way. In addition white dwarfs are size ground. (the size of Redmiond?) https://docs.google.com/Doc?docid=0AQIg8QuzTONQZGZxenF2NnNfNzY4ZDRxcnJ0aHI&hl=en_GB This is a good study of Arabic - English (scientific). Not a great text but illustrative. Irony is is that if I put in "Astronomy" and go though all the nouns in the passage I get a series of references from Google search which contains far better information than the original text. Translation clearly is very close to AGI. LATENT SEMANTIC ANALYSIS Google uses LSA/I to index its documents. It clearly does not in GT. Again What is the difference between constructing a dam in temperatures of up to 50C and fighting Israel? Google translated "وسط جو المعركة " as "central air battle". Correct is "the climatic environmental battle" or a more free translation would be "the battle against climate and environment". It should be noted that this is Sadat's speech on the completion of the Aswan High Dam. Clearly any use of LSA would tell you that you were NOT fighting Israel. In addition there have been some really terrible politically charged translations. I am sure you are aware of these. The fact of the matter is that there is no form of checking in "suggest better translation". Any translation "better" or not has to translate what is in front of it. https://sites.google.com/site/aitranslationproject/Home/software/examplefiles/sin is an example of assigning different meanings to words. Google does this solely by reference to bilingual text. I feel certain that Google is able to take all of this on board. What seems to be required is better liaison between Google Translate and other Google teams particularly in the general search area. What is needed is :- 1) The assignment of probabilities to translation via LSA. 2) Better parsing and assignment of pronouns (Elle).. 3) Check for reasonableness in better translations. 4) Feedback to search data as in the Stefan Boltzmann law. I would be interested to see what Google's comments are. - Ian Parker On Aug 9, 3:59 am, Xi Cheng (Google employee) wrote: > the ultimate goal is that the translation quality is so good that you > don't need to remember any practice like this. > in general, simple direct language, especially for the formal > expressions, work the best. that's only my experience. > > thanks for your feedback. > > On Aug 6, 11:27 pm, drtrueblue wrote: > > > > > Is there a good article or discussion that identifies the best > > practice for using this tool. For example, avoid expressions, idioms, > > literary syntax, etc. -- You received this message because you are subscribed to the Google Groups "General" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/google-translate-general?hl=en.
