Online Translators - Machine Translation

Do online translators really work, or are they more trouble than they're worth?

Machine translation refers to any kind of automatic translation, including translation software, hand-held translators, and online translators. While machine translation is an interesting concept and considerably cheaper and faster than professional translators, the reality is that machine translation is extremely poor in quality. Language is simply too complicated for machines to understand all of the vocabulary, grammar, context, and nuances in the source and target languages. The technology is improving, but the fact is that machine translation will never offer more than a general idea about what a text says. When it comes to translation, a machine simply cannot take the place of a human.

Online translators, which can be used to translate webpages, emails, or a pasted-in block of text, are no exception. To demonstrate some of the problems inherent in machine translation, I've run three sentences through the top five online translators. In order to check the accuracy, I then ran each translation back through the same translator (reverse translation is a common verification technique of professional translators). I've also provided my own translation for each sentence. Keep reading to find out how the online translators held up.

Remember that translation is an inexact science - there are always numerous acceptable possibilities.

Sentence 1: I love you very much, honey.

This is a very simple sentence - beginning students could translate it with little difficulty.

Online Translator	Translation	Reverse Translation
Systran	Je t'aime beaucoup, miel.	I love you much, honey.
Reverso	Je vous aime beaucoup, le miel.	I like very much you, the honey.
FreeTranslation	Je vous aime beaucoup, le miel.	I like you a lot, the honey.
InterTran	Je vous aime beaucoup , miel.	I love you a lot , honey.
Google Translate	Je t'aime beaucoup, le miel.	I love you very much, honey.

All five automatic translators took the word "honey" literally, rather than as a term of endearment, and three of them compounded the error by adding the definite article. The same three translated "you" as vous, which doesn't make much sense, given the meaning of the sentence. Reverso did particularly badly with its reverse translation - the word order is atrocious.

My translation: Je t'aime beaucoup, chéri.

Sentence 2: How many times did he tell you to write it?

Let's see if a subordinate clause causes any trouble.

Online Translator	Translation	Reverse Translation
Systran	Combien de fois vous a-t-il dit de lui écrire?	How much time does it have says you to write to him?
Reverso	Combien de fois vous a-t-il dit de l'écrire ?	How many times did he tell you to write it?
FreeTranslation	Combien de fois a-t-il dit que vous écrivez il?	How many times he says that you write it?
InterTran	Que de nombreuses fois a-t-il ajouté dit vous à écrivons le?	As many times has t - he added tell you at are writing him?
Google Translate	Combien de fois a-t-il de vous dire à l'écrire?	How many times did he tell you to write?

Systran inexplicably decided that "it" was an indirect object, rather than the direct object it is, which completely changed the meaning. In its reverse translation, it mistakenly translated the passé composé's auxiliary verb and main verb separately. Google added the preposition de, which makes it sound like "how many times does he have to tell you to write it," and in its reverse translation is lost the direct object. FreeTranslation did even worse, with a grammatically incorrect French translation. And InterTran's translation is completely bizarre. I have no idea why the nous form of écrire showed up, and I particularly like how in the reverse translation, it couldn't figure out what to do with the t- inversion. Reverso's translation and reverse translation are both excellent.

My translation: Combien de fois est-ce qu'il t'a dit de l'écrire ? or Combien de fois t'a-t-il dit de l'écrire ?

Sentence 3: Every summer, I drive up to the lake house and cruise around with my friends.

A longer and more complicated sentence.

All five translators were fooled by the phrasal verb "cruise around" and all but Google by "drive up" - they translated the verb and preposition separately. The pairing "house and cruise" caused problems - InterTran is the only one that knew that cruise was a verb rather than a noun, but it had a number of other serious problems: it didn't conjugate être and it had a terrible time with "my friends" - it didn't make mon plural to agree with amis, and then in the reverse it incorrectly translated mon as "mine," not to mention amis into Spanish! Google was fooled by et into thinking that I drive to the house and to the lake. Less shocking but still incorrect is the five programs' translation of drive as conduire - the latter is a transitive verb, but drive is used here intransitively.

My translation: Chaque été, je vais en voiture à la maison de lac et je roule avec mes amis.

Common Problems in Machine Translation

Though a small sample, the above translations offer a pretty good idea of the problems inherent in machine translation. The inability to use context is obviously the biggest flaw of machine translation - automatic translators are unable to use surrounding language (not to mention simple logic) in order to choose the correct translation of a noun with several meanings, as in the case of honey => miel/chéri. Excessively literal translations are common, and incorrect grammar plays a role too.

All in all, online translators can give you some idea about the meaning of a sentence, but their numerous flaws make it impossible for them to ever replace professional translators. If you're just after the gist and don't mind decoding the results, you can probably get by with an online translator. But if you need a translation that you can count on, hire a translator. What you lose in money you'll more than make up for in professionalism, accuracy, and dependability.