Why is translation so difficult?

To start off, I want to say I'm not a programmer.

That being said, why is translation so difficult? What's so hard about taking a phrase in one language and translating it to another?

Is Google translate getting any better?

Other urls found in this thread:

youtube.com/watch?v=QYlVJlmjLEc
en.wikipedia.org/wiki/List_of_linguistic_example_sentences
twitter.com/NSFWRedditImage

you need to take hrt as a teen

>you need to take hrt as a teen
I'm already taking hrt but I don't see how that's relevant

Learn another language, then you'll see how it's not so cut and dry. That's the most simple answer. If every language was the same with a different set of sounds what'd be the point of even having different languages at that point? Word order, multiple word meanings between regions, or even just readings, it's not as simple as input x get y.

Seriously, go read up on some other languages if you think it's so easy...

People do not talk in words, they talk in phrases. And while programs have no trouble translating single words, a phrase is constructed according to rules which vary greatly depending on the context. Said context often can be contained outside the picked phrase, in a previous sentence, or in a following one, in the greater written body containing it, in the intentions of the talker, even as far as in the head of the reader. These rules are by no means strict and are often even hard to formalize, wile a machine deals in formalized and strict algorithms. That does not mean that translators did not evolve through the past years, but they are still far from perfect. Consider the fact that even human beings struggle from time to time with translating texts that they poorly understand the context of. To create a perfect translator program we woul probably need to make it a functional AI. And even then its capabilities would be limited by its understanding of people. And people don't even understand each other that well.

Computers can't into the abstract. Now take your non-tech question and your fake nigger and piss off to Reddit.

If you literally ever learned a single language besides English, the answer would be obvious.

...

post feet

cultural expression get in language making translation be cognitive and cultural problem inside easy grammar rules + Dictionary.

Go break a leg, son! I have no beef with you, so let us bury the hatchet. Or are you still on the fence?

....translate that.

Context. Implied meaning. Subtle connotation. One sentence influencing the intended meaning of another much later on.

The sheer scope and complexity of a given language and the effort used to completely map that complexity to another language.

Keep in mind languages aren't word substitutions.

English has no word for lagom or orkar or boendestöd or many other swedish words. Swedish has no words for many things in English. And these are very similar languages.

How do I though in Swedish translate "my grandmother" from English.

In English it ambiguously refers to two possible people. In Swedish it is two words which each refer to a specific person.

And so on and so on

Lycka till gubbe! Jag har ingen problem med dig längre. Vad säger du om vi ska förlåta varandra.

And then whatever the last part is. Good luck teaching a computer to interpret this.

Good luck, man! I don't have any problems with you anymore. Let's say we leave each other?

Not great, but gives me a general enough idea.
Something like "bury the hatchet" is where machine translation falls down

So we lost the broken leg, the beef, the hatchet, and the fence.

do you even know two languages?

I understand Swedish though. But is that a metaphor or something?

The intermediary of translation by quine is a good paper to read to start ubdetstabdibg. There's been a few objections to the paper but it's the best starting poiny.

tfw humans can translate 'with an intensive porpoise' to its real meaning but no translation algorithm on the planet can even come close to figuring it out dynamically.

machine translation works well between the older noble languages, but that rules out english, which is only a couple hundred years old, and by design does not conform to the normal patterns of human communication that evolved naturally.
English was created for the commoners of England, to Isolate them from their brethren across the channel, to stop an uprising against the monarchy. As many a king and queen had been beheaded at the time It was a wise move on her majesty's part. Not so good for the people.

computers don't know kontext. google translate is fine for words or simple sentences, but good luck trying to translate a german sentence that uses formal you (Sie) to english and than translating it back to german pajeet

>why is translation so difficult?
Because languages are complex and often fluid, but translators are not.

Does she have purple nips?

youtube.com/watch?v=QYlVJlmjLEc

this might help you

basically because language is not just different ways of saying the same words, that would be fucking easy

these are terrible answers. MUH Context. MUH Ambiguity. Goes to show that you guys only have a entry level knowledge of linguistics and philosophy of language. The answer is indeterminacy

What the fuck are you talking about? English wasn't "created", it was brought to Britain by the Anglo-Saxons, which later blended with the language Viking invaders, and further on evolved through the Norman Conquest.

If you're referring to indeterminacy of translation, you are wrong, because translations of code can be demonstrated to be equivalent to one another, despite having different internal languages of expression. Computer programming languages are NOT natural languages, as the result of the computation can and must be quantified.

In fact, you could not have chosen an example of indeterminacy of translation that is more inappropriate and incorrect, due to the difference between the languages such indeterminacy was originally applied to (natural language) and the determinacy of languages you are trying to apply it to (quantity in, quantity out computer programs).

Human language is too ambiguous to parse with a well-defined set of rules and requires context.

en.wikipedia.org/wiki/List_of_linguistic_example_sentences

I've been learning German for about two years now. The words are easy and similar to English. The grammar is an absolute fucking nightmare. If you translated German literally to English, it would sound like an autistic lord of the rings larper with turrets.

There you go.
/thread

Google translate used to use a phrase based system, that basically translates individual words and phrases based on statistics. This is really primitive and requires buttloads of data. Part of the reason it doesn't work well is that languages have completely different grammar.

Now they are using very advanced recurrent neural networks with attention to do translation. One NN reads a sentences and translate it to an "intermediate language" which is just a bunch of numbers. Then another NN translates that into a sentence in the desired language. The NNs can learn to rearrange words to properly match the grammar rules of a language, which phrased based systems suck at.

And it can pick out context, like inferring a person is male in a language with gender neutral pronouns, and using "he" when translating to English.

Pick related, Google translate is rapidly approaching human quality translations on some languages. And even in the worst cases, it's score is still closer to human quality than it is to the old system.

Are you legally retarded?

Because programs aren't any good at understanding the meaning of languages.

>I'm going to work
substitute work with 仕事
>I'm going out
substitute out with 出る
>I'm going to work out
substitute work with 仕事 and out with 出る (incorrect)

Lmao.