Why is Google translate Thai-English so incredibly bad?

[Jingthing]

I don't think there is any doubt that the Thai-English google translator is shockingly poor. I wonder why exactly. I can compare it to Spanish-English which I would expect to be better, but not so MUCH better.

I assume a large part of the problem is the nature of the Thai language. Is that the only reason? Is google's tool not as good as other automatic Thai-English translation tools?

[DavidHouston]

I would pose the question the exact opposite way around: "Why does Google translate does as good a job as it does do under the circumstances?" What are the circumstances?

First Google uses the methodology called "statistical machine-translation". The following is from WIkipedia:

"Google Translate is a free, multilingual statistical machine-translation service provided by Google Inc. to translate written text from one language into another."

See http://en.wikipedia.org/wiki/Google_Translate and the section called "Translation Methodology". The statistical method is explained as follows: "According to Och, a solid base for developing a usable statistical machine translation system for a new pair of languages from scratch would consist of a bilingual text corpus (or parallel collection) of more than a million words, and two monolingual corpora each of more than a billion words.^[27] Statistical models from these data are then used to translate between those languages."

Google's difficulties are summarized, "Google Translate, like other automatic translation tools, has its limitations. While it can help the reader to understand the general content of a foreign language text, it does not always deliver accurate translations. Some languages produce better results than others. Google Translate performs well especially when English is the target language and the source language is one of the languages of the European Union. Results of analyses were reported in 2010, showing that French to English translation is relatively accurate^[9] and 2011 and 2012 showing that Italian to English translation is relatively accurate as well.^[10][11] However, rule-based machine translations perform better if the text to be translated is shorter; this effect is particularly evident in Chinese to English translations.^[9]"

English-Spanish-English has a huge database of translated material from which to draw statistical inferences; the amount of Thai-English-Thai material available in the Google database is necessarily much more limited. This limitation means that accuracy necessarily suffers.

[thaicbr]

And also how people write things. Sometimes I ask my gf to translate something and it takes time because of how they have worded it. Other times they are using slang.

Sent from my i-mobile i-STYLE Q6

[jpinx]

Google does a word by word translation but sentence construction is totally different in Thai so it comes out almost unintelligible until you work out where the subject, verb and object are. So easy to make mistakes because of this, but if you know basic thai word order and put the english into that, the answer is usually understandable to a thai person.

[JimMagee]

Forget Google translate for Thai to English. It works OK for English to Thai. I use http://www.thai2english.com for translating Thai to English. It works pretty good - much better than Google, plus it provides the tranliterated text as well. Useful for also learning Thai if you study the individual translationsl

[Cuban]

Part of the problem is that people that use English - use it badly when attempting to construct sentences that machine translators can cope with.

Often they use English as it is spoken - which is bizarre to many non-native ears when doing literal translations of words - often where the word translated is out of context. Or a slang or euphemism is used.

To get the best results.

You need to clearly understand what you need to say.

And use short sentences with no ambiguity

When you use the word right do you mean opposite of left or correct?

Try these ideas and use another Google Translation Tab to reverse the text and improve on the choice of words that you use.

The quality of the result depends on the effort you expend.

In 1970s computing we had the acronym GIGO...

...Garbage In Garbage Out.

[HeavyDrinker]

Sorry lads...hard yards with Mary Haas is the only way.....

Ain't no easy way out.....

[kriswillems]

I think there are many reasons:

- the difference between a verb and an adjective is not always very clear in Thai. In English there's a very clear difference in writing, and you can't mix them.

For instance: "a big dog" and "the dog is big" are to different sentence(parts) in English with 2 different meanings. But in Thai หมาตัวใหญ่ can mean both, depending on the context.

- Another problem is the word ได้. This single word can have many functions depending on it's position in the sentence but also the context. It can indicate the past time or an ability and a few other things. If google fails to find statistical information on the sentence, it will do word by word translation, which will fail.

- In Thai it hard to say where a word stops or starts because the are no spaces between words. There are algorithms to detect with a certain probability where a word stops and starts, but to be 100% sure you sometimes need to understand the context.

- In Thai it's hard to find out where a sentence stops. Several block with spaces between can sometimes be translated as only one sentence in English. Again, the algorithm for finding how many "blocks" of the Thai text you've to take into account to build one English sentence is very complicated and also here sometimes an understanding of the context is necessary. Because statistical database contains very few long sentences, google will fail on translating complicated constructions (because it falls back on word by word translation).

- It very often necessary to have an understanding of Thai culture to make good translations. European cultures, traditions and customs don't differ that much.

- Thai uses particles to express feelings (while we use intonation). The particles are hard to translate. Further on the same particle can express different feelings depending on the context. It hard for a computer to understand the context.

- European languages are very similar in sentence construction. If you fall back on word to word translation because there's no statistical data available, the translation is still understandable, because the order of words in several European languages is similar. In Thai the order of words can be very different.

- Almost every word in one European language has a perfect translation in the other languages. Thai words sometimes need to be explained (there's no word for them in English). Also one Thai word can have several meaning depending on the context. In European languages different meanings of words "overlap" much more.

Edited September 5, 2013 by kriswillems

[cheesyd]

You need a corpus. For decades & in some case centuries, loads of government documents, novels, research papers etc etc have been expertly translated into multiple European languages by private companies and governments. So the translate algorithms have lots of info to work with. With other languages there isnt always this massive replication of professionally translated, identical documents. The Thai - English corpus is minute compared to , say , German - English.

So you get the mumbo jumbo that is Thai - English translation on Google. It also accounts for the fact that Google is pretty good when it comes to translating well written, grammatically correct sentences but when you start throwing in slang it gets confused.

( This from a Google engineer who I know. )

[desi]

Does anyone know how much weight the crowdsourced corrections carry? I've been running the same phrases (plus a few added each year) through GT for five years now and the results don't seem to be improving that much. Could it just be a matter of not enough people interested in correcting Thai?

[Just1Voice]

Something I have found amusing using Google is that it will translate almost perfectly anything of a sexual context. I tried this just for experimental purposes only

Also, sometimes when asking questions, you have to "reverse" the sentence in some ways to get it to come out correctly.

When translating English to Thai, I then copy/past the Thai and reverse the translation to make sure it says what I want.

When I do use it, I keep sentences short, no paragraphs, using simple words, and that usually works fairly good.

[wpcoe]

I always get a chuckle when Google Translate turns "Hua Hin" to "San Francisco." Not sure exactly what context triggers that, but when copy and pasting from Thai web sites, I see it often.

[Just1Voice]

I always get a chuckle when Google Translate turns "Hua Hin" to "San Francisco." Not sure exactly what context triggers that, but when copy and pasting from Thai web sites, I see it often.

Or Bangkok, Chiang Mai, Khon Kaen, and almost any other city name/province, into "London".

[HeavyDrinker]

Kris deserves kudos for explaining this far better than I could or most could bar Richard W who is never around when you need him...

[Artist]

I am on Google+ a lot of the time and I get comments in many different languages. It does these incredibly well. This morning I was looking at some of the comments my wife and her friends were saying in Thai in Facebook. The language translation was by Bing and it was horrible just garbage. So I tried the same translation in Google and I think it was slightly better than Bing but still useless. No wonder I find it so difficult to understand my wife!

[HeavyDrinker]

Artist

I'm afraid there is no 'quick fix'.

There are some tools which do decent word for word trans, but the only way to really 'get there' is to get out the "dictionyelly" plus the grammar books and hack it out laddie....(recommend David Smyth's Thai grammar book thingymybob)

[BKKdreaming]

put something in Google translate and then Bing !

Bing is the translator for facebook ,

Anyway its like 2 stories that have a little in common but not much !

[Morakot]

Sorry lads...hard yards with Mary Haas is the only way.....

Ain't no easy way out.....

This is a good example of another important issue!

The less the source text conforms to formal/ standard linguistic pattern (in terms of grammar, style, and idiomatic expression) the lower will be accuracy of the target language.

Edited September 5, 2013 by Morakot

[brewsterbudgen]

Forget Google translate for Thai to English. It works OK for English to Thai. I use http://www.thai2english.com for translating Thai to English. It works pretty good - much better than Google, plus it provides the tranliterated text as well. Useful for also learning Thai if you study the individual translationsl

Yes, this is the best one for a quick online translation.

[GirlDrinkDrunk]

it is a damn sight better than Bing

[desi]

Google Translate vrs T2E ... it all depends what the phrase is.

Put แล้วเจอกัน in both. T2E breaks it down. GT simply gives "see you".

Same goes with แล้วคุณล่ะ - T2E breaks it down word by word. GT has "How about you."

พูดภาษาไทยได้นิดหน่อย - T2E comes into its own (a range of possibilities is helpful with longer sentences especially). GT fails with "Thailand speaks a little."

[Fookhaht]

Count your blessings. The English-to-Thai is even worse.

[atyclb]

A good way is to look up the individual thai words in a sentence and then try to figure it out. This also helps to learn new words and improve thai reading ability.

http://dict.longdo.com/ is another resource

Edited September 5, 2013 by atyclb

[desi]

I prefer learning phrases because word for word in Thai doesn't always make sense. I like to see the breakdown though (I'm driven to see the breakdown - can't help myself).

[desi]

cant or can't?

[Johnnie99]

Learn how to use it properly and it is fine.

A bad workman blames his tools.

[Fookhaht]

Learn how to use it properly and it is fine.

A bad workman blames his tools.

Cultural-translation into Thai, please! "A bad dancer blames the drummer."

[newermonkey]

Facebook (Bing) is just as useless as Google at translating from Thai to English, In fact it should say "Translate to gibberish"

"Babylon" (free translator gives you 5 translations only!!!) is a bit better, at least you can get the rough idea of what is being said in Thai, but they want a subscription every month for the non free one.

Edited September 5, 2013 by newermonkey

[duanebigsby]

Forget Google translate for Thai to English. It works OK for English to Thai. I use http://www.thai2english.com for translating Thai to English. It works pretty good - much better than Google, plus it provides the tranliterated text as well. Useful for also learning Thai if you study the individual translationsl

That's what I use as well. I double check it against google translate and between the two I can understand most of it.

Two things confuse the translation, slang and local dialects. I'm in Chiang Rai, and no translator can handle my Thai friends Facebook post when they use Thai Lanna.

[kriswillems]

I think the remark about slang language is a very good remark. In Thai writing language or formal language and speaking language are very different. When I just finished my first Thai course in BKK I could understand the prime minister when he would speak on TV, but I could not understand any of the workers that came to build/fix our house (even when they spoke central Thai). In school they taught me formal language.

Because the corpus is probably based on writing language or formal language google translate is really useless when we try to translate the messages our new girlfriend posts or facebook or her chat/e-mail messages we spy on