Sometimes I usage Google Language Tools or various other translation sites to check sentences that I"ve developed on my own. Recently I"ve tried copying muzic-ivan.info sentences out of textbooks right into these sites and noticed the the translation often tends to be way off.

You are watching: Why is google translate so bad at japanese

Is this bereason muzic-ivan.info can be hard to translate literally? Or indicative of the top quality of these tools?


*

MatthewD is correct--computer translations are unmost likely to be very precise, bereason computers aren"t exceptionally good at understanding herbal language yet. Think about it--as soon as you use the word "fair" in a sentence, for example, a listener has to figure out if you intend "simply," "mediocre," "reasonable," or also "festival." Sometimes tbelow isn"t enough indevelopment for also a fluent speaker to figure out what you suppose, and computer systems are far from fluent speakers at the moment.

I would certainly prefer to add that muzic-ivan.infoEnglish translations are particularly vulnerable to inaccuracy, for a pair of reasons:

muzic-ivan.info sentences are much more most likely to be ambiguous (from an English speaker"s perspective), just as English sentences are likely to be overly-specific (from a muzic-ivan.info speaker"s perspective).

muzic-ivan.info has actually the majority of words that don"t conveniently analyze into English, favor まま, はず, こと, and the assorted collection phrases like いただきます and よろしくおねがいします.


Share
Follow
answered Jun 24 "11 at 0:23
*

Amanda SAmanda S
7,7671010 silver badges1414 bronze badges
3
Add a comment |
9
Google Translate is horrible. Due to its statistical methods, it makes absurd mistakes that not even other machine translators would ever before make. I"ve viewed it do points like change the subject from "you" to "we", even in languperiods wright here the subject is entirely unambiguous in both languages. Even the other machine translators aren"t that great, either, for reasons explained in the other answers.

This is my favorite example of machine mistranslation. I wanted to recognize if tright here was a Spanish term tantamount to "hardball" -- i.e., tough and also unforproviding. I put in "It"s a hardsphere human being." I gained earlier, "Es un muncarry out del béisbol" -- "It"s a civilization of baseround."


Share
Follow
answered Jun 24 "11 at 2:21
*

Kef SchecterKef Schecter
2,4301010 silver badges22 bronze badges
2
Add a comment |
9
Not really sure this question drops into the scope of JLU. And 100% certain a thostormy answer wouldn"t.

Answer: No, you can not.

Machine translation (MT) is a very old trouble, people have actually been at it for the previous 60 years, and also while algorithms have actually gone a lengthy means over the previous 10 years, they are still miles ameans from "trustworthy" in the language pair you are interested in.

Btw, contrary to what some people have composed above, symbolic MT (which counts on "understanding" the text in order to interpret it) does not develop excellent results and also is presently not the favoured technique. After being touted as the Divine Grail of MT (and among the major goal of AI research) in the 1980s, it"s been mostly discarded in useful applications nowadays. Many effective contemporary units (Google Translate etc) use a heavily statistical approach: they manage word connections and sentence frameworks in an abstract method, through little-to-no attempt at understanding the definition behind them.

See more: Functional Category That Has The Greatest Amount Of Movement

This may seem counter-intuitive, but it functions extremely well for languperiods that are reasonably cshed together, linguistically speaking (even on very ambiguous phrasing): EnglishFrenchSpanish and so on is fast getting to commercial high quality.

Unfortunately for you, Indo-European and Japonic langueras could not be sitting any kind of farther on the great language tree (wildly different syntax, dealing with of referential and all sorts of etymological terms I am not familiar enough to offer you). Which implies it will probably take one more decade or so before symbolic MT or (more realistically) advanced statistical models offers any kind of usable result on muzic-ivan.info English.