Google translator

The text in the box is moved from my other blog!
Looking back I should have given a better title like: an analysis of Google translator

I was talking with a friend and at some time he told me a German sentence – I was impressed, because it was grammatically correct and also the pronunciation was good. Later he told a 2nd sentence – and again I was impressed (but this time the pronunciation was not so good :-) ). When I asked about the German sentences, he told he used Google translator. I told that Google translator must be good.
But the next 3 translated sentences proved me wrong – I could not get the meaning at all. Also not, when I translated the sentence word for word back to English (reason: I assumed Google would translate it like this). So I had to take back my initial assumption.
But did I have enough data to say something at all on the quality?

So, we agreed, that I do some further testing on this.

How did I proceed?

1. Search some background info on Google translator (e.g. it was fed with 200 billion words).
2. But well, I did not want to read at that time too much and also maintain my curiosity, so I continued like this:
   get me some random sentences.
   
   I went to yahoo and searched for “english site” and took this website for getting me some random English text. Reason: this is a site especially for learning and teaching English – as it tells: “A web-site for teachers and learners of English as a secondary language from a German point of view”.
   So, my intention here was: to get something for the benefit for Google, as I assume this kind of text would be better translated (do not get me wrong: for testing I would take later some text which is e.g. more technical, nested, … to challenge the application more :-) )
3. Then I decided to take about every 5th sentence from there (why every 5th? well, which would you have used?) - the 16 sentences are listed in below table.
4. So, I started to translate these sentences with Google translator to German (since I have no version info on Google Translator for rechecking later: at 20.3.2007, 21:38 German time (GMT+1)). The sentences which (from my point of view) were translated perfect, I marked with OK - see below table.
   So this makes until now a ratio of about 48% (2 out of the 5 sentences from the conversation and 8 from the above 16 sentences).

Then I translated the remaining 8 not-OK sentences how I would like them myself (using one time LEO) - see below table again.

Actually the tests are not yet finished, I just gave some random input (16 sentences only - from my point of view easy sentences) to get me an impression on the quality. So, we have now a number - 48% - whatever this might tell. I wonder how much other online translators (Babelfish uses Systran, PROMT, … ) would achieve? But does this prove anything so far? It just tells:
for these sentences it had this result - how good this works for each one of us is another case.

But well, it is late and perhaps you can help me testing? Otherwise I will continue soon.

I searched (of course in Google :-) ) for other old bugs from Google translator. Found e.g. these: hit, hit, hit

BTW:
Michael has listed here web services for English/German + vice versa translations. My all time favorite is LEO.
Michael plans to write a series on translation tools - so drop by his blog.

input list: 16 sentences

number	sentence	step4 results	my translation
1	He has given his stick a name.	OK
2	As you can see, the Auxiliary and the subject change their places.		Wie du sehen kannst, tauschen das Hilfsverb und das Subjekt ihre Plätze.
3	This boy will come running if Mr. Verb forgets his stick and takes him by his hand.		Dieser Junge kommt angerannt, wenn Herr Verb seinen Stock vergißt und nimmt ihn an seiner Hand.
4	But in a question there must be something in front of the subject.		Aber in einer Frage muß es etwas vor dem Subjekt geben.
5	That’s the one thing, which he buys.	OK
6	The same thing happens in the simple past.	OK
7	And at last here’s the sentence diagram of the English question once more.	OK
8	I cannot play the piano.		Ich kann das Klavier nicht spielen.
9	They went swimming.		Sie gingen schwimmen.
10	First you have to put the sentences in the negative form and afterward you’ll have to find questions.		Zuerst mußt du die Sätze in die negative Form setzen und danach mußt du Fragen finden.
11	We have gone shopping		Wir sind Einkaufen gegangen.
12	We go shopping every day.		Wir gehen täglich einkaufen.
13	Helga likes milk.	OK
14	You must get up at 7 o’clock.	OK
15	I went to bed at 8 o’clock.	OK
16	Betty and Peter speak English

Author: Erkan Yilmaz (2007-03-20)

comments

• I'm using these days mainly: Google Translator, LEO + dict.cc,

see also

• Blog:Observations/eine Analyse von Google Translator (German version of this posting)
• category:Language

notes

— Erkan Yilmaz (talk) • 01:19, 20 March 2007 • 21 comments