Google Translate just nearly doubled its number of supported languages

AusPeter

Ars Tribunus Militum
2,780
Subscriptor
Google translate is a good resource, not a great one. Just yesterday I was reviewing some Spanish and it totally did not understand my grammatically correct usage of the present subjunctive, and kept wanting to change my text to present indicative. Fortunately I know Spanish well enough that I didn't fall for the trap.
 
Upvote
6 (33 / -27)

Flailsafe

Smack-Fu Master, in training
65
Google translate is a good resource, not a great one. ...
I'd say Google Translate is a great resource, just not a perfect one. But "perfection" in regards to language are not compatible bedfellows.

It's pretty high on my "Please-Google-Don't-Kill-This" list
 
Upvote
136 (136 / 0)
Post content hidden for low score. Show…

BirdOcean

Smack-Fu Master, in training
13
It says they used PaLM 2 to support these languages, so they're using AI to do the translation.

That could be a good thing or a bad thing depending on how well it does. I've done some tests with Google Translate, and ChatGPT-4o, between English and Japanese, and I've noticed that ChatGPT-4o came up with translations that were far more natural sounding and seemed more likely, to me.

But I also know that ChatGPT-3.5 and 4 (so far I don't yet have enough experience with 4o to know for sure) have made weird mistakes with translations as well, totally rewriting stuff. So it can be a crapshoot.

I hope Google's project to use AI to translate languages rather than their previous method turns out to be more accurate. But I guess we'll find out when people upload screenshots of its weird screwups and total rewrites... or don't upload those screenshots if it turns out they don't happen.
 
Upvote
13 (18 / -5)
Post content hidden for low score. Show…

AusPeter

Ars Tribunus Militum
2,780
Subscriptor
I'd say Google Translate is a great resource, just not a perfect one. But "perfection" in regards to language are not compatible bedfellows.

It's pretty high on my "Please-Google-Don't-Kill-This" list
The problem is that google translate can introduce errors in text that people who can't manually translate will miss. And if they can manually translate it, then they don't really need google.

And I'm talking about simple and obvious grammatical structures, not fancy stuff.
 
Upvote
6 (11 / -5)

marsilies

Ars Legatus Legionis
21,948
Subscriptor++
It says they used PaLM 2 to support these languages, so they're using AI to do the translation.

That could be a good thing or a bad thing depending on how well it does.
Technically, Google Translate has been using AI since 2016.

In 2016, Google Translate gradually replaced the older statistical machine translation approach with the newer neural-networks-based approach that included a seq2seq model combined by LSTM and the "additive" kind of attention mechanism. They achieved a higher level of performance than the statistical approach, which took ten years to develop, in only nine months.


Current LLMs basically evolved from AI translation methods.
The big breakthrough was a model from Google called "the transformer." The researchers at Google were working on a very specific natural language problem: translation...

Transformers were a breakthrough for translation, but they were also exactly the right model for solving many language problems.

They were perfect for working with GPUs because they could process big chunks of words in parallel instead of one at a time. Moreover, the transformer is a model that takes in one ordered sequence of symbols—in this case, words (technically fragments of words, called "tokens")—and then spits out another ordered sequence: words in another language.

And translation doesn’t require complicated labeling of the data. You simply give the computer input text in one language and output text in another. You can even train the model to fill in the blanks to guess what comes next if it's fed a particular sequence of text. This lets the model learn all kinds of patterns without requiring explicit labeling.

Of course, you don’t have to have English as the input and Japanese as the output. You can also translate between English and English! Think about many of the common language AI tasks, like summarizing a long essay into a few short paragraphs, reading a customer’s review of a product and deciding if it was positive or negative, or even something as complex as taking a story prompt and turning it into a compelling essay. These problems can all be structured as translating one chunk of English to another.

The big breakthrough in language models, in other words, was discovering an amazing model for translation and then figuring out how to turn general language tasks into translation problems.

So one would hope that since they started as translators, they can still do translation well.
 
Upvote
62 (62 / 0)
so they're using AI to do the translation.

That could be a good thing or a bad thing depending on how well it does.
Google Translate has been based on machine learning since 2016 and specifically on deep learning transformer architecture since 2020. PaLM2 is an evolution of the existing model used by Google Translate for the past few years.
 
Upvote
38 (38 / 0)
IMHO Google Translate, especially live translation of written text on smartphones and live 2-way translation of spoken conversations, is the craziest futuristic thing in my life. It is what I show those haters who like to say that Google hasn't innovated since GMail (or whatever their reference year may be).
 
Upvote
55 (56 / -1)

phaedrus11

Smack-Fu Master, in training
77
Subscriptor++
IMHO Google Translate, especially live translation of written text on smartphones and live 2-way translation of spoken conversations, is the craziest futuristic thing in my life. It is what I show those haters who like to say that Google hasn't innovated since GMail (or whatever their reference year may be).
I use the live translation of written text almost every day. It's pretty awesome. I use it to translate medication labels and menus.
 
Upvote
14 (14 / 0)

marsilies

Ars Legatus Legionis
21,948
Subscriptor++
How come translate.google.com still only lists 128 languages?
I count 245 in Chrome on Desktop:

1719523425042.png

My Android app hasn't updated yet though.

Edit: List of languages in text:
Abkhaz
Acehnese
Acholi
Afar
Afrikaans
Albanian
Alur
Amharic
Arabic
Armenian
Assamese
Avar
Awadhi
Aymara
Azerbaijani
Balinese
Baluchi
Bambara
Baoulé
Bashkir
Basque
Batak Karo
Batak Simalungun
Batak Toba
Belarusian
Bemba
Bengali
Betawi
Bhojpuri
Bikol
Bosnian
Breton
Bulgarian
Buryat
Cantonese
Catalan
Cebuano
Chamorro
Chechen
Chichewa
Chinese (Simplified)
history
Chinese (Traditional)
Chuukese
Chuvash
Corsican
Crimean Tatar
Croatian
Czech
Danish
Dari
Dhivehi
Dinka
Dogri
Dombe
Dutch
Dyula
Dzongkha
check
English
Esperanto
Estonian
Ewe
Faroese
Fijian
Filipino
Finnish
Fon
French
Frisian
Friulian
Fulani
Ga
Galician
Georgian
German
Greek
Guarani
Gujarati
Haitian Creole
Hakha Chin
Hausa
Hawaiian
Hebrew
Hiligaynon
Hindi
Hmong
Hungarian
Hunsrik
Iban
Icelandic
Igbo
Ilocano
Indonesian
Irish
Italian
Jamaican Patois
Japanese
Javanese
Jingpo
Kalaallisut
Kannada
Kanuri
Kapampangan
Kazakh
Khasi
Khmer
Kiga
Kikongo
Kinyarwanda
Kituba
Kokborok
Komi
Konkani
Korean
Krio
Kurdish (Kurmanji)
Kurdish (Sorani)
Kyrgyz
Lao
Latgalian
Latin
Latvian
Ligurian
Limburgish
Lingala
Lithuanian
Lombard
Luganda
Luo
Luxembourgish
Macedonian
Madurese
Maithili
Makassar
Malagasy
Malay
Malay (Jawi)
Malayalam
Maltese
Mam
Manx
Maori
Marathi
Marshallese
Marwadi
Mauritian Creole
Meadow Mari
Meiteilon (Manipuri)
Minang
Mizo
Mongolian
Myanmar (Burmese)
Nahuatl (Eastern Huasteca)
Ndau
Ndebele (South)
Nepalbhasa (Newari)
Nepali
NKo
Norwegian
Nuer
Occitan
Odia (Oriya)
Oromo
Ossetian
Pangasinan
Papiamento
Pashto
Persian
Polish
Portuguese (Brazil)
Portuguese (Portugal)
Punjabi (Gurmukhi)
Punjabi (Shahmukhi)
Quechua
Qʼeqchiʼ
Romani
Romanian
Rundi
Russian
Sami (North)
Samoan
Sango
Sanskrit
Santali
Scots Gaelic
Sepedi
Serbian
Sesotho
Seychellois Creole
Shan
Shona
Sicilian
Silesian
Sindhi
Sinhala
Slovak
Slovenian
Somali
Spanish
Sundanese
Susu
Swahili
Swati
Swedish
Tahitian
Tajik
Tamazight
Tamazight (Tifinagh)
Tamil
Tatar
Telugu
Tetum
Thai
Tibetan
Tigrinya
Tiv
Tok Pisin
Tongan
Tsonga
Tswana
Tulu
Tumbuka
Turkish
Turkmen
Tuvan
Twi
Udmurt
Ukrainian
Urdu
Uyghur
Uzbek
Venda
Venetian
Vietnamese
Waray
Welsh
Wolof
Xhosa
Yakut
Yiddish
Yoruba
Yucatec Maya
Zapotec
Zulu
 
Last edited:
Upvote
13 (13 / 0)

adespoton

Ars Tribunus Angusticlavius
8,873
I count 245 in Chrome on Desktop:

View attachment 84105

My Android app hasn't updated yet though.

Edit: List of languages in text:
Abkhaz
Acehnese
Acholi
Afar
Afrikaans
Albanian
Alur
Amharic
Arabic
Armenian
Assamese
Avar
Awadhi
Aymara
Azerbaijani
Balinese
Baluchi
Bambara
Baoulé
Bashkir
Basque
Batak Karo
Batak Simalungun
Batak Toba
Belarusian
Bemba
Bengali
Betawi
Bhojpuri
Bikol
Bosnian
Breton
Bulgarian
Buryat
Cantonese
Catalan
Cebuano
Chamorro
Chechen
Chichewa
Chinese (Simplified)
history
Chinese (Traditional)
Chuukese
Chuvash
Corsican
Crimean Tatar
Croatian
Czech
Danish
Dari
Dhivehi
Dinka
Dogri
Dombe
Dutch
Dyula
Dzongkha
check
English
Esperanto
Estonian
Ewe
Faroese
Fijian
Filipino
Finnish
Fon
French
Frisian
Friulian
Fulani
Ga
Galician
Georgian
German
Greek
Guarani
Gujarati
Haitian Creole
Hakha Chin
Hausa
Hawaiian
Hebrew
Hiligaynon
Hindi
Hmong
Hungarian
Hunsrik
Iban
Icelandic
Igbo
Ilocano
Indonesian
Irish
Italian
Jamaican Patois
Japanese
Javanese
Jingpo
Kalaallisut
Kannada
Kanuri
Kapampangan
Kazakh
Khasi
Khmer
Kiga
Kikongo
Kinyarwanda
Kituba
Kokborok
Komi
Konkani
Korean
Krio
Kurdish (Kurmanji)
Kurdish (Sorani)
Kyrgyz
Lao
Latgalian
Latin
Latvian
Ligurian
Limburgish
Lingala
Lithuanian
Lombard
Luganda
Luo
Luxembourgish
Macedonian
Madurese
Maithili
Makassar
Malagasy
Malay
Malay (Jawi)
Malayalam
Maltese
Mam
Manx
Maori
Marathi
Marshallese
Marwadi
Mauritian Creole
Meadow Mari
Meiteilon (Manipuri)
Minang
Mizo
Mongolian
Myanmar (Burmese)
Nahuatl (Eastern Huasteca)
Ndau
Ndebele (South)
Nepalbhasa (Newari)
Nepali
NKo
Norwegian
Nuer
Occitan
Odia (Oriya)
Oromo
Ossetian
Pangasinan
Papiamento
Pashto
Persian
Polish
Portuguese (Brazil)
Portuguese (Portugal)
Punjabi (Gurmukhi)
Punjabi (Shahmukhi)
Quechua
Qʼeqchiʼ
Romani
Romanian
Rundi
Russian
Sami (North)
Samoan
Sango
Sanskrit
Santali
Scots Gaelic
Sepedi
Serbian
Sesotho
Seychellois Creole
Shan
Shona
Sicilian
Silesian
Sindhi
Sinhala
Slovak
Slovenian
Somali
Spanish
Sundanese
Susu
Swahili
Swati
Swedish
Tahitian
Tajik
Tamazight
Tamazight (Tifinagh)
Tamil
Tatar
Telugu
Tetum
Thai
Tibetan
Tigrinya
Tiv
Tok Pisin
Tongan
Tsonga
Tswana
Tulu
Tumbuka
Turkish
Turkmen
Tuvan
Twi
Udmurt
Ukrainian
Urdu
Uyghur
Uzbek
Venda
Venetian
Vietnamese
Waray
Welsh
Wolof
Xhosa
Yakut
Yiddish
Yoruba
Yucatec Maya
Zapotec
Zulu
Well that's interesting; Edge isn't showing that many for me -- I wonder if it's the browser or my location that affects the number.
 
Upvote
5 (7 / -2)

marsilies

Ars Legatus Legionis
21,948
Subscriptor++
Well that's interesting; Edge isn't showing that many for me -- I wonder if it's the browser or my location that affects the number.
Well, the Google Blog post says they're "rolling out" the update. So that means it may be available to some users/browsers, but not others.

I checked Edge on my PC, and that has the shorter list. But then I thought maybe it's because I'm not logged in into Google in Edge, whereas I am logged in to Google in Chrome, so I opened an Incognito window in Chrome, and that has the shorter list too.

So they seem to be rolling it out to Google logged-in users first.
 
Last edited:
Upvote
18 (18 / 0)
Well that's interesting; Edge isn't showing that many for me -- I wonder if it's the browser or my location that affects the number.
Every feature at Google rolls out slowly to a random subset of users that slowly ramps up to 100% so they can track whether it's causing problems. It's extremely normal for one person to see a feature and another person not see it.
 
Upvote
22 (22 / 0)

DanNeely

Ars Legatus Legionis
15,401
Subscriptor
Are they also working on improving language translations they've supported for years but do a very poor job of? Chinese or Japanese to English translations tend to be barely intelligible and frequently get me/you and yes/no reversed. From clarification discussions (and these are so much fun when you can't trust the translations to be correct if there isn't someone bilingual to sort the mess out) translations from English to Chinese or Japanese are similarly awful.
 
Upvote
7 (9 / -2)

keltor

Ars Praefectus
5,546
Subscriptor
Are they also working on improving language translations they've supported for years but do a very poor job of? Chinese or Japanese to English translations tend to be barely intelligible and frequently get me/you and yes/no reversed. From clarification discussions (and these are so much fun when you can't trust the translations to be correct if there isn't someone bilingual to sort the mess out) translations from English to Chinese or Japanese are similarly awful.
Mandarin to English or Japanese to English present some of the actual worst case scenarios that exist. Even people like my kids who are native speakers in both Japanese and English will struggle to translate without entirely rewriting the content (which is what happens in quality professional translations.)


Chat GPT4 is kind of interesting in its translations, it can do some amazing things with translations that are one of the things I find impressive, but then it will mistranslate simple stuff.
 
Upvote
23 (23 / 0)
Mandarin to English or Japanese to English present some of the actual worst case scenarios that exist. Even people like my kids who are native speakers in both Japanese and English will struggle to translate without entirely rewriting the content (which is what happens in quality professional translations.)


Chat GPT4 is kind of interesting in its translations, it can do some amazing things with translations that are one of the things I find impressive, but then it will mistranslate simple stuff.
Wow that’s interesting. So what makes that pair so bad?
 
Upvote
2 (2 / 0)
When I saw Manx in the headline, I had to laugh. As a student of Irish, Google Translate is of... limited use. It's gotten to the point of academic humor, with folks like Gearóidín McEvoy commenting on Motherfoclóir (fun podcast about the Irish language) that she can instantly tell which of her students used it because the errors are glaring and consistent. It's gotten better in the last five or so years, but it's a looong way from the quality of Spanish, French, and German on the same tool.
 
Last edited:
Upvote
6 (8 / -2)
I had to learn two languages last century for my doctorate because I had to be able to do proper research and at least somewhat communicate in them. Some of my peers had to learn three. One old friend, had to learn classical Greek and her thesis was gnarly, and her advisor required fluency. In ancient Greek.

Keep in mind, this was necessary because we would teach the next generation. It's an unbroken chain without a shortcut. Only one chapter requires research in language X? Too bad.

It's strange to think that so much of those old rigors are rapidly becoming obsolete, almost in the blink of an eye historically speaking. Language is a huge and ancient hurdle. It doesn't really exist anymore.
 
Upvote
9 (10 / -1)

pfstevenson32

Wise, Aged Ars Veteran
125
Subscriptor
In Google Translate of Mandarin, which I am trying to learn, you cannot enter the diacritics of pinyin. (Well, you can, but you get nothing back.) So, you enter the words using an english keyboard and, once the whole sentence is entered, it usually is in the right ball park. Trying to get a word, which is usually 2 syllables, is pretty hit or miss. It just seems so backward. (I am nowhere near good enough with spoken Mandarin to start learning the characters, beyond the easy ones.)
 
Upvote
2 (3 / -1)
I had to learn two languages last century for my doctorate because I had to be able to do proper research and at least somewhat communicate in them. Some of my peers had to learn three. One old friend, had to learn classical Greek and her thesis was gnarly, and her advisor required fluency. In ancient Greek.

Keep in mind, this was necessary because we would teach the next generation. It's an unbroken chain without a shortcut. Only one chapter requires research in language X? Too bad.

It's strange to think that so much of those old rigors are rapidly becoming obsolete, almost in the blink of an eye historically speaking. Language is a huge and ancient hurdle. It doesn't really exist anymore.
Here's what happens when you round trip that comment via Chinese (Simplified):

google said:
In the last century, in order to obtain a doctorate, I had to learn two languages, because I had to be able to conduct appropriate research and at least to use them to communicate. Some of my classmates have to learn three -door language. An old friend had to learn classical Greek. Her papers were difficult. Her instructor asked fluently to master ancient Greek.

Remember, this is necessary because we have to teach the next generation. This is an uninterrupted chain without shortcuts. Only one chapter needs to study a certain language? so terrible.

It is strange that from a historical point of view, these ancient rigor is rapidly outdated, almost in the blink of an eye. Language is a huge and ancient obstacle. It does not exist anymore.

More than meets the eye.
 
Upvote
7 (8 / -1)

JanneM

Ars Scholae Palatinae
604
Subscriptor++
I'd say Google Translate is a great resource, just not a perfect one. But "perfection" in regards to language are not compatible bedfellows.

It's pretty high on my "Please-Google-Don't-Kill-This" list

It really depends on which languages you translate in between. For similar languages - between any Indo-European languages for instance - it does a credible job.

Between very dissimilar languages it fares much worse. I've occasionally used it between japanese and english and the results vary from clunky to utterly incomprehensible.

It seems to depend on how much context you need to capture the same ideas. Similar languges share a lot of structure, and you can focus on the sentence level or even smaller. But for dissimilar ones you may need to look at - and rewrite - a paragraph or two at a time, to get a coherent translation. Google translate frequently fails at that.
 
Upvote
6 (7 / -1)

lucubratory

Ars Scholae Palatinae
1,178
Subscriptor++
In Google Translate of Mandarin, which I am trying to learn, you cannot enter the diacritics of pinyin. (Well, you can, but you get nothing back.) So, you enter the words using an english keyboard and, once the whole sentence is entered, it usually is in the right ball park. Trying to get a word, which is usually 2 syllables, is pretty hit or miss. It just seems so backward. (I am nowhere near good enough with spoken Mandarin to start learning the characters, beyond the easy ones.)
Hello from another Mandarin-learner 👋 You will need to use a software Chinese keyboard if you want to learn Chinese, even though you can long press characters to add tones on the English keyboard. On Android the Simplified Chinese keyboard is 非常好, excellent. The way a Chinese software keyboard works (native Chinese speakers use this too) is that you have a pinyin keyboard, and it automatically translates your pinyin into a selection of characters to choose from using its knowledge of the language. The default selection is normally correct but in two years of using it I have never encountered a situation where the correct words weren't at least on the list after I input my pinyin. When using the Google translate app on Android, the keyboard will automatically switch depending on which field you're entering into, so it works seamlessly to write Chinese into the Chinese field and English into the English field. You can overwrite this choice manually if needed. To activate the pinyin keyboard go to anywhere the keyboard pops up, click on the upper left of the keyboard that has the four boxes symbol, then click settings, languages, add keyboard, and add “中文(简体)”。If you need to know exactly where to click there are video guides on YouTube that should help; pick a recent one.


If you are using Google translate primarily on desktop, you will need to use a desktop software Chinese keyboard. On Windows, go to Start, Settings, Time & Language, Language & Region, Add a language, and select "Chinese (Simplified, China)". Once you have added that language, click on the three little dots to the right of "Chinese (Simplified, China)" and select "Language options". Select "Add a keyboard", then choose Microsoft Pinyin. After that is done, select/tick every option on this page aside from Handwriting character conversion in order to download all the components you might need in the future. When you need to type in Chinese, simply press the Windows key + spacebar to switch keyboards, or select the Chinese keyboard from the taskbar on the right. You can activate or deactivate the keyboard seamlessly while using Google Translate on desktop, and the keyboard works the same as the Android one: type in Pinyin, a selection of words will come up, press a number to indicate which word you are trying to write (or touch it, or mouse click it).


A side note that I'm not sure is relevant, but may be helpful: Literacy in Chinese is assessed by knowing the Chinese language's characters; once you know the characters and appropriate grammar (very easy in Chinese), you can write most sentences in Chinese. Pinyin is more analogous to an input method or a pronunciation guide; it is not the actual language and knowing pinyin in the absence of knowing the characters cannot result in literacy. It's totally possible to learn to speak and hear Chinese without learning the characters; it is known as illiteracy and is fortunately nearly eradicated in China itself. I would personally not recommend intentionally waiting to learn characters until you are more proficient at speaking and hearing - coming back from illiteracy can be very difficult and demoralising, and learning characters complements learning spoken and written Chinese, it doesn't interfere with it at all. Characters can be intimidating but they are a lot easier once you've learned enough of them to start recognising radicals. If it helps you, you may also wish to learn stroke order/handwriting to learn characters, but in my opinion this is not mandatory to learn Chinese - it will be a knowledge gap between you and native Chinese speakers, but is irrelevant for almost all electronic communications. I would put learning chéngyǔ/成语 after a few years of studying Chinese at a much higher priority than learning stroke order.


Good luck with learning Chinese!
 
Upvote
31 (31 / 0)