Google Translate

[techno_race]

It appears someone at Google has gotten bored lately.
January 24, 2002:
1772
January 5, 2005:
1773
March 15, 2007:
1774

[techno_race]

May 14, 2008:
1775

[djr33]

I'm skeptical of that. I expect that they all run through English, and since English is so broken, you'll end up with really weird stuff.

EDIT: It's absolutely going through English [tested it], which is a terrible idea. First time I've strongly disagreed with google, but that's really idiotic. Word for word translations from Italian to Spanish will VERY often be better than anything with today's publicly available technology put through English in the middle. Nonsense.

Now, on the other hand, it's good that you can so easily translate between two languages that aren't English, thus saving yourself one step of conversion, but it's no better than the results we could have had before this new input menu was constructed. It's really misleading as well.

Example:
Spanish: Me llamo Daniel.
Expected Italian: Mi chiamo Daniel.
[English translation: My name is Daniel.]
Resulting Italian: Il mio nome è Daniel.
And that's just not what someone would say in Italian-- it should match the Spanish exactly.

The much better idea would be to not fix it for English-- rather, just translate to broken English, then go from there. But I guess that assumes the languages are related. If you went from Chinese to Russian, and through English, this method would probably give better results.

[HostV]

From what I heard Google wants to perfect this within few years so they can translate any webpages on the net and interchangeably. Hope to see this happen! :)

[Master_script_maker]

as you can see they are very close:
http://dynamicdrive.com/forums/attachment.php?attachmentid=1842&d=1213817027

[molendijk]

From what I heard Google wants to perfect this within few years so they can translate any webpages on the net and interchangeably. Hope to see this happen! :)That is totally impossible. Even if Google were able to resolve all the syntactical problems pertaining to translating (sequences of) sentences from one language into another language, they would face an enormous range of other types of problems, among which metaphors (taken in their broadest sense). For instance, in Dutch you can say 'he took the legs' to refer to someone who got away quickly. In French, you would say 'he took the legs into his neck' in these circumstances. In India, people say 'she walks like an elephant' to indicate the graciousness of the walking, whereas English speakers would prefer 'she walks like a gazelle' (or something like that; I'm not sure), etc.
===
Arie Molendijk

[djr33]

Translation is always imperfect; however, it would be possible to get a literal translation to work properly.
On the other hand, using AI it would be possible to design a learning computer that would function as the human mind does, perhaps reading books in each language to learn. Certainly nothing simple, though.

[hyk]

i agree with djr33
Translation is always imperfect You might not get the exact meaning/ emotion across to the receiver if you translate it.

good luck to google in translating the pages then, perhaps they could prove us wrong

[Twey]

On the other hand, using AI it would be possible to design a learning computer that would function as the human mind does, perhaps reading books in each language to learn. Certainly nothing simple, though.Interestingly, that is exactly what Google is attempting to do (and how it differs from other translation solutions on the market, such as Apartium or SysTram). Google Translate is fed large amounts of bilingual text, with correspondences in meaning indicated to it, and stores the frequency of each translation, using this to guess the most appropriate translation.

I strongly doubt that they use English as an intermediate language; if they do, it would only be as a fall-back, when no direct translation were available.

[djr33]

I strongly doubt that they use English as an intermediate language; if they do, it would only be as a fall-back, when no direct translation were available.No... I'm positive they do. That's how it goes from one to the other... through English. It used to not be available from one to another, because they didn't have the algorithms. I fully believe they are trying to slowly replace this with real language to language parsers, in time, though.

[Twey]

As I said: as a fallback, because it has a lack of sufficient literature in those languages to for it to analyse the direct translation.

[molendijk]

Translation is always imperfect; however, it would be possible to get a literal translation to work properly.
Literal translations can give you hilarious text. If I literally translate Dutch 'hij nam de benen' into French 'il a pris les jambes' (he took the legs), then a native speaker of French will interpret the sentence as referring to someone taking the legs of somebody else (his wife's legs, for instance). This is just an example among thousands of similar cases.

On the other hand, using AI it would be possible to design a learning computer that would function as the human mind does, perhaps reading books in each language to learn. Certainly nothing simple, though.
Not simple indeed. At the level of the sentence, AI may be a possibility. At the level of sequences of sentences, things will get extremely complicated (I would say: impossible to handle with AI). Interpreting sequences of sentences is very much a complicated question of combining syntactical and semantic knowledge with world knowledge (and world knowledge is definitely infinite!!). How could we know, with the help of AI (only), that English 'John fell. Max pushed him' is interpreted as 'Max pushed John, then John fell'', whereas the French literal translation of the sequence: 'Jean tomba. Max le poussa' means (default meaning) that John fell first and was pushed afterwards? (The difference between the English and French sequence is related, among other things, to the differences between the tenses of English and French)

Interestingly, that is exactly what Google is attempting to do (and how it differs from other translation solutions on the market, such as Apartium or SysTram). Google Translate is fed large amounts of bilingual text, with correspondences in meaning indicated to it, and stores the frequency of each translation, using this to guess the most appropriate translation.
Twey: what's the link with AI? Storing enormous amounts af linguistic data cannot be qualified as 'intelligent/intelligene'.
===
Arie

[Twey]

I don't think I mentioned AI...

[molendijk]

I don't think I mentioned AI...
I concluded it from this:

Quote (DJR):
On the other hand, using AI it would be possible to design a learning computer that would function as the human mind does, perhaps reading books in each language to learn. Certainly nothing simple, though.
Twey: Interestingly, that is exactly what Google is attempting to do (and how it differs from other translation solutions on the market, such as Apartium or SysTram). Google Translate is fed large amounts of bilingual text, with correspondences in meaning indicated to it, and stores the frequency of each translation, using this to guess the most appropriate translation.
===
Arie.

[djr33]

As I said: as a fallback, because it has a lack of sufficient literature in those languages to for it to analyse the direct translation.Well, that implies it to be an exception. At the moment, it's definitely NOT an exception, but the general method used. In time, they will work to make it only a fallback, but right now, "fallback" or not, it's how it translates everything. And it's not about how much literature is available, but how much has been done with this literature to make it work for translation.

[molendijk]

[...] At the moment, it's definitely NOT an exception, but the general method used.
So we would have:
Je vois l'homme (French) --> I see the man (English) --> Ich sehe de? Mann (German).
We should have ich sehe deN Mann, but as English doesn't have case (nominative, accusative etc.), this would be impossible to get. Is that Google's method, DJR? That would be most stupid.
===
Arie.

[djr33]

It's not that bad.

It's a translation from French to English, then from English to German. That doesn't mean it won't have any cases, but it means it WILL go through the middle step of English, which forces more lost data than going directly. It'll be a decent translation to English, then a decent translation into German, but it'll have the inevitable loss from two steps, and also not the help of translating between two languages that may be more similar to each other than to English.

Also, English is such a messed up language, it's probably the worst to use as a middle language (except that it's already programmed).



Looks like it's now not only going through English

Here's what we get for the French->German you posted above:
"Ich sehe den Menschen"

Now, if I run it through English, I first get:
"I see the man"
Then to German:
"Ich sehe den Mann"

How strange.

From "je m'appele Daniel" I get "Ich heisse Daniel" in German. Running through English it becomes "I'm Daniel", then "Ich bin Daniel" in German.

Well, that's great. They've finally stopped using English in the middle.

[molendijk]

DJR, that's an improvement on Google's side.
By the way, why do you think English is a broken and messed up language? Would that mean that French (for instance, or whatever other language) is superior to English IN MANY RESPECTS? If not, you cannot maintain your claim about English.
(There'a a lot of normal English 'things' that French cannot do. French cannot say, for instance, 'he walked to the kitchen'. It must use the equivalent of 'he went to the kitchen'. Adding how he got there ('walking') would make the sentence rather heavy. Just an example of what French cannot do).
===
Arie.

[djr33]

English is an awkward hybrid of Anglo-Saxan (old english, germanic language) and French, thus indirectly bringing in Latin and Greek. The biggest problems with this are shown in the horrible spelling system in English, but it's also present in other ways. It's a hybrid, thus has a lot of awkward features syntactically. English is moving toward being an isolating language (like Chinese), yet insists on holding on to some odd grammatical things, such as "whom", or more frequently "him".

English is not all that difficult to learn, true, but it's extremely difficult to master because it is entirely irregular. It also is not a very opportune language for translation because it doesn't use many structures that exist in other languages.

Thus, English is illogical and random-- it works, but only through memorization and not through much to do with patterns. This is shown in those learning English, with, for example, how difficult it can be to learn which preposition to use with which verb, sometimes making an entirely new meaning:
I want to look up the word.
I want to look the word up.
I get up from bed.
I get from bed up.


As for French, I do think it is superior to English in many ways, though specifically with French I have other complaints. I think that both German and Italian are better than English, though.

[hyk]

French sound nicer than English when it is spoken from my point of view. Generally I had not much complaints about English other than the tenses stuffs& sometimes the pronounction stuffs.

For example: present,past tense, future tenses,. :p I don`t really remember the names of these groups anymore. i learnt it in my primary school times. whereas in your country it should be junior high school ( correct me if I`m wrong).


djr33: Does French & German sound similar?

to put it plainly, I believe all major languages have their pros & cons. Maybe 1 of these days someone out there will create an even better language to use :D

[djr33]

French and German are completely different in sound. They do have a couple of the same sounds, though, like the R. But overall, German is very harsh and French is very soft.

Tenses are complex in many languages, though in English they're not very consistent and can be confusing. For example, you said "Does French & German sound similar", and it should be "DO French & German sound similar". But why? English is a mess. There is absolutely no reason for using "do" or "does" except grammar. I can give you a complex grammatical reason and even a historical analysis for why that works like that, but right now, it's just pointless. In Spanish, for example, you can just skip the subject, like "hablo" means I talk, and "hablas" means you talk, so you don't need to say "yo" (I) or "tu" (you). In English, though, we can't just say "Runs." We need to say "He runs." even though that is the only option, when there's an S.

Yes, all languages have pros and cons (not just the major ones). English has more problems, though, than other languages.

I think my favorite language is, as far as I know now, Italian, and with just a few changes it would be very regular and easy. As it is now, it's still very regular and easy, but there are a couple weird things.

As for creating languages, I'm interested in this, and I'm slowly working at making a language. But there are many other "constructed languages" (or conlangs) out there. The most popular is Esperanto, which is spoken by (some sources say) 2 million people. http://en.wikipedia.org/wiki/Esperanto

[molendijk]

DJR, I'm afraid I don't agree with certain statements you made.

1. HYBRID NATURE OF ENGLISH
English is a hybrid of Anglo-Saxon (old english, germanic language) and French, thus indirectly bringing in Latin and Greek. Problems with that: awkward spelling system and syntactic features.

All languages are naturally hybrid. As linguistic communication often takes place between groups speaking different languages, the communication tools (languages) influence each other. The 'awkwardness' of syntactic features resulting from that (your words, not mine) is not language-specific. The roots of French, for instance, were in the spoken version of Latin that Roman conquerors brought to foreign territory after their invasion of 'La Gaule' in 59 B.C. After the Roman empire fell in the 5th century, France was invaded by germanic tribes (Franks), who brought features of their own languages into Old French: French is hybrid as well.

2. SPELLING
That's about the only artificial thing of a language. It's not just the result of spontaneous linguistic evolution, but something imposed upon us by certain official groups (like spelling committees).

3. IRREGULARITIES
English is as (ir)regular as any language. The question of which preposition to use with which verb, for instance, is a difficult problem (to non-native speakers) in all languages. Just one example for French: être obligé DE (to be forced TO) vs obliger quelqu'un À (force someone TO).

4. FUNCTIONALITY OF USING CERTAIN FORMS
The fact that Spanish and Italian are 'pro(noun)-drop' languages, whereas English (and germanic languages in general) have to use pronouns does have reasons. In English, verbs hardly have conjugational characteristics (I walk, you walk, he walkS, we walk, you walk, they walk), whereas Spanish and Italian have a rich conjugational system (also much richer than French). So, from a functional point of view, it is not hard to understand why the latter languages can easily drop the (subjects-) pronoun.

5. CONSTRUCTED LANGUAGES
The fact that Esperanto has turned out to be a complete failure shows that languages are 'living things' that cannot be artificially imitated. If Esperanto would have succeeded in surviving (by itself), then it would have rapidly created its own irregularities. There's enough linguistic evidence for that.

6. WHAT IS GRAMMAR?
If a language like English insists on holding on to some 'odd' grammatical things, such as 'whom' or 'him', then it's because the language itself wants it, not because some 'grammarian' prescribes it. And why would 'him' be odd? In all languages I know of (not just in English), there's a very strong tendency to keep different pronominal forms for different cases (case: subject vs object), even if the case-system as such doesn't exist
anymore (like in Dutch, English, French ...).
===
Arie.

[Twey]

As for creating languages, I'm interested in this, and I'm slowly working at making a language. But there are many other "constructed languages" (or conlangs) out there. The most popular is Esperanto, which is spoken by (some sources say) 2 million people.There are a variety of languages created precisely for the purpose you describe. Most are patented and/or not designed to be spoken by humans, but suitability this kind of thing is one of the design goals of Lojban.
The fact that Esperanto has turned out to be a complete failure shows that languages are 'living things' that cannot be artificially imitated. If Esperanto would have succeeded in surviving (by itself), then it would have rapidly created its own irregularities. There's enough linguistic evidence for that.Hm? Who said Esperanto hasn't succeeded in surviving? It's still around and spoken today, widely if not densely. You can find Esperanto speakers and even clubs or groups in most major cities of the world. Also, people are now, I'm told, beginning to suspect that Latin may have been a constructed language, which nobody can call unsuccessful (even if you don't consider its descendants to be successes of the language itself).

[molendijk]

You can find Esperanto speakers and even clubs or groups in most major cities of the world.
Yes, Esperanto is (legitimately!) maintained by groups who want it to survive. But it never reached the stage of a language that is 'native' to a group. So it does not 'grow', it is simply kept alive.



Also, people are now, I'm told, beginning to suspect that Latin may have been a constructed language, which nobody can call unsuccessful (even if you don't consider its descendants to be successes of the language itself).
A large part of what we know about classical Latin comes from texts written by poets (Horace, Ovid ...), orators (Cicero. Tacitus ...). The 'ciselled' language they use is rather artificial, as is the language produced by Shakespeare. Taken in that sense, classical Latin is 'constructed'. But luckily, there are remains (graffiti) of spoken Latin ('latin vulgaire') that was just as natural (as a language) as English, French, German etc.
As for the descendants of Latin, they are children of spoken Latin, not of written Latin.
===
Arie.

[djr33]

1. HYBRID NATURE OF ENGLISH
[...]
All languages are naturally hybrid.All languages are a hybrid in some sense due to borrowing but I am confident in saying that English is, at least of the most popular languages in the world (and perhaps all, excluding pidgins, etc.) is the MOST mixed between two bases. Old French has influences, yes, but it's not a 50-50 mix of Latin/Greek and Old Germanic roots, like English is. Though English is a Germanic language (by origin and classification), a slight majority of the words are actually from Latin and/or Greek originally.


2. SPELLING
That's about the only artificial thing of a language. It's not just the result of spontaneous linguistic evolution, but something imposed upon us by certain official groups (like spelling committees).Yes, and in the case of English it's a screwed up mess. Spanish, German, Arabic and other languages aren't messed up like English.


3. IRREGULARITIES
English is as (ir)regular as any language.No. Citing specific examples is not a way to prove this, either. Every language has irregularities, but English has a lot, especially noted in the spelling. Do native speakers of German have trouble spelling? Sure, a bit. Do native speakers of Spanish have trouble spelling? Sure, a bit. Do native speakers of English have trouble spelling? Of course!


4. FUNCTIONALITY OF USING CERTAIN FORMS
The fact that Spanish and Italian are 'pro(noun)-drop' languages, whereas English (and germanic languages in general) have to use pronouns does have reasons.Right, but in those languages, then, the conjugations have a function. The S in the third person singular present tense conjugation of verbs in English is completely worthless (just left over from old germanic). I'm not saying there aren't differences-- but I am saying it's an awkward mess right now.


5. CONSTRUCTED LANGUAGES
The fact that Esperanto has turned out to be a complete failure[...]As Twey said, Esperanto is successful. You certainly cannot force a language upon a group, even with extreme political means-- language change and adaptation is guaranteed and NOTHING will stop it. And considering this, Esperanto is very successful. Plus, I'm not sure where your argument was really going, so feel free to expand if you'd like.


6. WHAT IS GRAMMAR?
If a language like English insists on holding on to some 'odd' grammatical things, such as 'whom' or 'him', then it's because the language itself wants it, not because some 'grammarian' prescribes it. And why would 'him' be odd? In all languages I know of (not just in English), there's a very strong tendency to keep different pronominal forms for different cases (case: subject vs object), even if the case-system as such doesn't exist
anymore (like in Dutch, English, French ...).1. It IS because the grammarian wants it-- people correct you when you turn in an academic paper without using 'whom'. Thus, it stays with us today. I personally use 'whom' to be correct, and I did not develop this naturally. It is not something developing because the language wants it, but rather falling OUT of use because the language does NOT want it.
2. Of the languages you cited, you're only basing it on the indo-european branch of languages-- there are many more, though this tendency does appear a lot.
English simply makes no use of this any more, though, because word order is so clear. And this is falling out of use:
"Joe was talking to Sam and I." That sounds totally normal, though it is wrong, according to what I hear every day.
German, Latin, and other languages make use of pronoun forms to say different things:
"Dich sehe ich." and "Ich sehe dich." mean the same thing.
"You see I" and "I see you" certainly don't mean the same thing, and get confusing. Additionally, "you" doesn't decline, showing our grammar no longer needs declining pronouns. The extraneous nature of "whom" proves this as well.

---Post 2----

Yes, Esperanto is (legitimately!) maintained by groups who want it to survive. But it never reached the stage of a language that is 'native' to a group. So it does not 'grow', it is simply kept alive.It DID grow. It is said to be spoken by up to 2 million people, and it was created about a century ago when no one spoke it. It is still growing today. It's also said now that some children are growing up as native speakers, with parents who only share Esperanto as a common language, though I haven't seen proof of this yet.


The 'ciselled' language they use is rather artificial, as is the language produced by Shakespeare.[....]As for the descendants of Latin, they are children of spoken Latin, not of written Latin.True.
Twey, I'm not quite sure what you're getting at about Latin being constructed, as it has been basically proven to be a descendant of PIE. But perhaps you meant this as molendijk says, that is it sculpted by the elite for what we consider to be 'classical' latin. I'd believe that, pending more evidence.
And I love Latin. It's such a simple language, though very different from English. I've got a book about Vulgar latin giving birth to the many dialects we now consider different languages I've been meaning to read. I should start that.

[Twey]

Yes, Esperanto is (legitimately!) maintained by groups who want it to survive. But it never reached the stage of a language that is 'native' to a group. So it does not 'grow', it is simply kept alive.As a language constructed with the sole purpose of being an IAL, it's never meant to replace natural languages.
Twey, I'm not quite sure what you're getting at about Latin being constructed, as it has been basically proven to be a descendant of PIE.That says nothing about how it come to be, though. Esperanto, for example, can be considered a descendant of fairly large group of languages, yet that is only so because the language has been constructed to be so.

[molendijk]

Daniel, you illustrate what you say about the hybrid nature of English with facts concerning the English vocabulary only. Now, the hybrid nature of the English vocabulary is not in itself a burden for speakers of that language. And it may sometimes be a blessing for non-native speakers (like speakers of French learning English). On a side note: you may be surprised by the heterogenuous vocabulary of French, see this (http://www.espacefrancais.com/vocabulaire/origine_mots.html).

As for spelling: there's a Dutch spelling committee that regularly 'modernizes' the spelling of Dutch words. The result? Confusion all over the place and loss of visible signs of the origin of the words. A modern phonetic spelling system is not necessarily better than a non-phonetic one.

You say that native speakers of English have a lot of trouble spelling. That's strange. I speak Dutch (my native langugae) and studied thoroughly (not just a holiday course) English, German, French, classical Latin and vulgar Latin (and a bit of Spanish and Italian). I cannot say that English language gave me more spelling trouble then the other languages. And that also goes for lots of other people I know.

Which languages do you think I found the toughest ones to learn and to master, generally? Latin and German.
And the easiest? English! No specific spelling problemsl.

About functionality: I agree that the S in 'he runS' seems to be superfluous. So one might expect it to disappear one day. However: natural language is not always 'logic', so you cannot tell. (A nice French example illustrating that: you say 'les deux premiers' ('the two first') instead of '*les premiers deux' ('the first two'), although there can only be one first).

What I said about the failure of Esperanto was ment to illustrate the impossibility of creating a language as a fullgrown replacement of a natural language. The majority of inhabitants of this planet learns English, not Esperanto, for communication purposes (like they learned Latin in the old days).

About grammar: if the average speaker of English is not bothered by 'Joe was talking to Sam and I' and normally uses it, then this only proves that grammar is not what grammarians prescribe, but what language speakers use / what language wants / how language evolves. If a colleague tells you to replace 'I' with 'me' in that sentence, then it's because he knows that certain people (purists and the like) won't take you
seriously if you use 'I'. But that has nothing to do with grammar, but rather with how people judge other people by their acts.

(I hope there aren't too many language errors in the above text (haha) and happily provide you this link: the news in classical Latin (http://www.yleradio1.fi/nuntii/)).
===
Arie.