| Pages in topic: < [1 2] | Machine Translation - A sentence to put technology to a test over time? Thread poster: Tomás Cano Binder, BA, CT
|
|---|
| Fixed by brute force | Feb 14, 2009 |
José Henrique Lamensdorf wrote:
Earlier machine translation used to render...
ACME utilizes the latest technology...
in Portuguese as:
A ACME utiliza a tecnologia mais atrasada...
Indeed!  I think these things will be fixed by force, not by intelligence... Google Translator has fixed this with brute force. Use it to translate "That does not happen in the latest version" and you will see "....en la versión más reciente" (correct) and not "...en la última version" (incorrect, but many people use it). So they must have added the "más reciente" thing to their databases at some stage.
| | | | | No syntax... but sheer processing power | Feb 14, 2009 |
Neil Coffey wrote:
As you can see, the system essentially knows no syntax.
Indeed! Your was a very good analysis. I sincerely thank you for that.
"Daniel said that flying planes is dangerous. Bob said that flying planes are dangerous." Good example. I think many examples like this will be easily fixed at some time based on free translations provided by non-translators in Google.
So indeed the goal here is to know whether the machine can "think" instead of "checking its archives".
| | | | | Indeed, no syntax | Feb 14, 2009 |
Spanish: "¿Cómo? ¿Cómo como? ¡Como como como! Como como como como, ¿cómo me lo preguntas?"
To the Spanish-speaker listener or reader, the accents and syntax make the meaning as clear as water (even if it sounds puzzling for a second). It says something like "What? How I eat? I eat the way I eat! Since I eat the way I eat, how come you ask me about it?"
Four MTs:
"How? How how? Like like like! As like as like, how am I you ask?"
"How? How as(lik... See more Spanish: "¿Cómo? ¿Cómo como? ¡Como como como! Como como como como, ¿cómo me lo preguntas?"
To the Spanish-speaker listener or reader, the accents and syntax make the meaning as clear as water (even if it sounds puzzling for a second). It says something like "What? How I eat? I eat the way I eat! Since I eat the way I eat, how come you ask me about it?"
Four MTs:
"How? How how? Like like like! As like as like, how am I you ask?"
"How? How as(like)? Like like like! Like like like like, how do you ask it me?"
"How? How ace (like)? Like like like! Like like like like, how do you ask it me?"
"How? How like? How how how! How how how how: how do you ask me about it?" (I was pleased with this one; at least it did grasp the final part!)
I now fear that this could be picked up by Google and be added to their database at some stage, thus making this forum useless. Should we exchange sentences over email, assign them a codename and evaluate new MT systems against them without expressing the translation in public anywhere? ▲ Collapse
| | | | | Protect die Wiese! | Feb 14, 2009 |
German: "Die die die die die Wiese betreten sehen dies anzeigen oder melden erhalten eine Belohnung".
Means something like "Anyone seeing anyone stepping onto the lawn and informing about it or reporting it will be rewarded".
Some MTs:
"See entering into the meadow this indicate or announce get a reward."
"Those those those those the meadow see this entered indicating or announce receive a reward"
"To the meadow to enter see this show or ... See more German: "Die die die die die Wiese betreten sehen dies anzeigen oder melden erhalten eine Belohnung".
Means something like "Anyone seeing anyone stepping onto the lawn and informing about it or reporting it will be rewarded".
Some MTs:
"See entering into the meadow this indicate or announce get a reward."
"Those those those those the meadow see this entered indicating or announce receive a reward"
"To the meadow to enter see this show or report will receive a reward"
"The meadow see entering this register or announce receive a reward." ▲ Collapse
| | |
|
|
|
Samuel Murray 
Netherlands
Local time: 16:20
Member (2006)
English to Afrikaans
+ ...
| Deliberately confusing sentences | Feb 14, 2009 |
Tomás Cano Binder, CT wrote:
Spanish: "¿Cómo? ¿Cómo como? ¡Como como como! Como como como como, ¿cómo me lo preguntas?"
German: "Die die die die die Wiese betreten sehen dies anzeigen oder melden erhalten eine Belohnung".
You can have a lot of fun with such sentences, but I don't think any good author will use them in a published text except for amusement or illustrative value. These are trick sentences, and their purpose is not to communicate.
If you want to be "honest" or "fair" in your difficult sentences for MT, you should use sentences that are likely to occur in a newspaper or magazine or in a novel that doesn't score high on the artsy-fartsy novelty scale.
The sentence pair about the planes is a good one.
I suspect translating English into Esperanto may also be a good test, with the subject/object confusion, although I don't know of any EN-ESP MT systems.
| | | | | garden-path sentences | Feb 14, 2009 |
I believe the so-called garden-path sentences are hard to process:
1. The horse raced past the barn fell.
2. The man who hunts ducks out on weekends.
3. The cotton clothing is usually made of grows in Mississippi.
4. The prime number few.
5. Fat people eat accumulates.
6. The tycoon sold the offshore oil tracts for a lot of money wanted to kill JR.
(taken from Pinker 1994)
I do not know how MT works but if yo... See more I believe the so-called garden-path sentences are hard to process:
1. The horse raced past the barn fell.
2. The man who hunts ducks out on weekends.
3. The cotton clothing is usually made of grows in Mississippi.
4. The prime number few.
5. Fat people eat accumulates.
6. The tycoon sold the offshore oil tracts for a lot of money wanted to kill JR.
(taken from Pinker 1994)
I do not know how MT works but if you liked the "flying airplanes" then I guess you will like the above sentences as well. Garden-path sentences are hard for humans to process (you must have experienced that), so I guess MT will have difficulties too.
Lilia ▲ Collapse
| | | |
Evangelia Mouma wrote:
1. The horse raced past the barn fell.
2. The man who hunts ducks out on weekends.
3. The cotton clothing is usually made of grows in Mississippi.
4. The prime number few.
5. Fat people eat accumulates.
6. The tycoon sold the offshore oil tracts for a lot of money wanted to kill JR.
Very good ones! Thanks for sharing them here.
Indeed, garden-path signs are the hardest. Specially German ones!
How about something in Greek? Can you think of something we could use?
| | | | | subject-objects | Feb 14, 2009 |
In Greek every word has an ending that shows its role in the sentence (subject - object etc). So, I believe that it is difficult to construct garden-path sentences. On the other hand, since there are endings, we can put the words anywhere in the sentence (following of course the rules of constituents) with slight difference in meaning; in other words, there is more or less free word order. Now, I say again that I do not know how MT works but if we give it a sentence where there is no indication ... See more In Greek every word has an ending that shows its role in the sentence (subject - object etc). So, I believe that it is difficult to construct garden-path sentences. On the other hand, since there are endings, we can put the words anywhere in the sentence (following of course the rules of constituents) with slight difference in meaning; in other words, there is more or less free word order. Now, I say again that I do not know how MT works but if we give it a sentence where there is no indication from the ending of the noun about its role in the sentence, then the MT won't manage to translate it. Example:
to sholio misise to pedi (by the way, sholio is pronounced s-h-olio, that is, s and h separately)
(word-to-word translation)
the school hated the child
(translation)
the child hated (going to) school
Here is the sentence in Greek: Το σχολείο μίσησε το παιδί
Here you need semantics in order to understand who hates who, since it is not shown by the endings (to sholio and to pedi are the same in nominative/accusative).
But then again, there shouldn't be many instances of "to sholio misise" in Google's database.
Anyway, I feel I am treading into fields I do not know well.
Lilia ▲ Collapse
| | |
|
|
|
Neil Coffey
United Kingdom
Local time: 15:20
French to English
+ ...
| Not necessarily brute force | Feb 14, 2009 |
Tomás Cano Binder, CT wrote: Indeed! I think these things will be fixed by force, not by intelligence... Google Translator has fixed this with brute force. Use it to translate "That does not happen in the latest version" and you will see "....en la versión más reciente" (correct) and not "...en la última version" (incorrect, but many people use it). So they must have added the "más reciente" thing to their databases at some stage.
Actually, Google Translate probably doesn't use brute force as such. What they do is expand their training data (so there are now enough instances of the phrase "versión más reciente" for it to be "picked up"), and improve their algorithm (so e.g. instead of looking at statistics of 3-word sequences, it looks at 5-word sequences).
By the way, if people are interested in more about how Google Translate (and indeed some of their text-processing technology in general) works, you may like to check out Peter Norvig's presentation on 'Theorizing from Data'.
| | | | Neil Coffey
United Kingdom
Local time: 15:20
French to English
+ ...
| Machine learning | Feb 14, 2009 |
So indeed the goal here is to know whether the machine can "think" instead of "checking its archives".
Actually, there's not such a clear distinction between the two concepts. The Machline Learning technique used by a system such as Google Translate is modelling "thinking" in the Artificial Intelligence sense of the word.
| | | | | Pages in topic: < [1 2] | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Machine Translation - A sentence to put technology to a test over time? | Wordfast Pro | Translation Memory Software for Any Platform
Exclusive discount for ProZ.com users!
Save over 13% when purchasing Wordfast Pro through ProZ.com. Wordfast is the world's #1 provider of platform-independent Translation Memory software. Consistently ranked the most user-friendly and highest value
Buy now! » |
| | TM-Town | Manage your TMs and Terms ... and boost your translation business
Are you ready for something fresh in the industry? TM-Town is a unique new site for you -- the freelance translator -- to store, manage and share translation memories (TMs) and glossaries...and potentially meet new clients on the basis of your prior work.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
