Add Transliteration (Romanization) for Translation API [35900102]

Feature Request

Status Update

No update yet.

Description

ju...@google.com

created issue #1

Dec 11, 2015 09:49PM

* What is the desired behavior of the feature?
-> Support pinyin transliteration or pronunciation representation for Chinese translation output at Google Translate API.

* If relevant, why are current approaches or workarounds insufficient?
-> Not Applicable, but

https://translate.google.com/ has the option of the pronunciation representation, please see the screenshot attached.

* If relevant, what new use cases will this feature will enable?
-> For names, such as Chinese names, Cities names, and other names, it will be more efficient and professional that these names will be auto-translated into pronunciation representation in a more appropriate way in terms of preciseness of translations.

* Similar Feature Request at google-ajax-apis:
->

https://code.google.com/p/google-ajax-apis/issues/detail?id=478&q=label%3AAPIType-Language&sort=-stars&colspec=ID%20Type%20Stars%20Status%20Modified%20Summary%20APIType%20Opened

GuGe.jpg

42 KB

View

Download

Comments

ju...@google.com <ju...@google.com> Apr 5, 2017 07:52PM

Status: New

ca...@gmail.com <ca...@gmail.com> #2Jun 16, 2018 08:35AM

Google Translate web provides this function, like

https://translate.google.com/#zh-CN/en/%E4%BB%8A%E5%A4%A9%E6%88%91%E4%BB%AC%E9%83%BD%E6%98%AF%E4%B8%AD%E5%9B%BD%E4%BA%BA

It converts "今天我们都是中国人" to "Today we are all Chinese", and you also can get "Jīntiān wǒmen dōu shì zhōngguó rén" under the text edit for source.

Chinese characters and Pinyin romanization are different writing systems or scripts for Chinese. The conversion between them can be considered as translation in some meaning. Simplified Chinese and Traditional Chinese are also different writing systems or scripts for Chinese.

I would like to see the feature implemented, thanks.

az...@mlvegankitchen.com <az...@mlvegankitchen.com> #3Sep 2, 2018 03:09AM

Would also like this feature very much

bo...@gmail.com <bo...@gmail.com> #4Oct 30, 2018 09:52AM

+1 On this, it'd be a great feature to have.

bo...@gmail.com <bo...@gmail.com> #5Jan 23, 2019 12:29AM

I am a software engineer working on cross linguistic parsing. I don't speak any Chinese. For the project I am working on, I am investigating how my parsing routines work with Chinese compared to English (as well as other languages). The resources that I use for understanding Chinese have examples with glosses in pinyin and English. (For example, Li, Thompson, 1981. Chinese Mandarin. A functional reference grammar). The pinyin glosses help the Chinese characters stick in the brain a little bit more, and thus helping with my overall effectiveness.

I use the translate API to see what a Chinese sentence is about, and also what the individual words in the sentence mean. If there was an API for converting Chinese words to pinyin, I would translate the test sentences to pinyin and display the logo-grams, pinyin and English all in parallel.

Thank you for your consideration.
Bob New

bo...@gmail.com <bo...@gmail.com> #6Feb 2, 2019 03:28AM

Follow up on my earlier use case for a pinyin translator.

I found the Chinese EDictionary. See this site: ttps://

cc-cedict.org/wiki/

The main file contains >100,000 entries with traditional, simplified, pinyin, and english translations. The format of the file is very easy to reverse engineer - I wrote a reader for the file pretty quickly. This will server my purpose, so I no longer am requesting google API to support translation to pinyin.

Thanks,
Bob New

al...@gmail.com <al...@gmail.com> #7Mar 18, 2019 10:35AM

Hey, I've built an app to learn languages and I'd also like to request this feature.. For beginners, it's not very useful to work with characters, but the Pinyin romanised versions could be really helpful!

Would be very very nice!

mk...@gmail.com <mk...@gmail.com> #8Jun 3, 2019 01:34AM

Hi All,

I normally use CEDICT for pinyin, but if a word does not exist in CEDICT (which is very often), then I would like to use a Google API.

Please add this feature.

Best regards,

Matthew

de...@gmail.com <de...@gmail.com> #9Feb 2, 2020 05:52PM

Also looking for this integration.

Kind regards,
Jeroen

di...@newskylabs.net <di...@newskylabs.net> #10Feb 15, 2020 07:21PM

Hi,

The same wish here.

I wonder why this is built into the google translate gui page - but not into the API:

When I enter 我是中国人 into

https://translate.google.com/ I get both - the pinyin romanization and the translation:

- I am Chinese
- Wǒ shì zhōngguó rén

Try here:

https://translate.google.com/#view=home&op=translate&sl=zh-CN&tl=en&text=我是中国人

But the query
service.translations().list(source='zh', target='en', q=['我是中国人']).execute()
only returns the translation:
{'translations': [{'translatedText': 'I am Chinese'}]}
rather than:
{'translations': [{'pinyin': 'Wǒ shì zhōngguó rén', 'translation': 'I am Chinese'}]}

I also tried the Google Cloud Natural Language API using the script

https://github.com/GoogleCloudPlatform/python-docs-samples/blob/master/language/api/analyze.py. But neither of

analyze.py entities 我是中国人
analyze.py sentiment 我是中国人
analyze.py syntax 我是中国人

returned any information about the pinyin romanization.

Is there anybody who has an idea how the google translate gui page implements the pinyin romanization - or how a similar result could be obtained?

Thanks for your help,

Dietrich

di...@newskylabs.net <di...@newskylabs.net> #11Feb 15, 2020 08:20PM

Answering to my own message:

> Is there anybody who has an idea how the google translate gui page implements the pinyin romanization - or how a similar result could be obtained?

One approach could be to first use the syntactical analysis of the Cloud Natural Language API (

https://cloud.google.com/natural-language/) to analyse the sentence into lexical units:

我是中国人

=>

nsubj root suff attr
我是中国人
PRON VERB NOUN NOUN

and to then feed these units into a dictionary providing pinyin romanization - for example the CC-CEDICT dictionary (

https://cc-cedict.org/wiki/).

This strategy would result in:

我 [wo3] /I/me/my/
是 [shi4] /is/are/am/yes/to be/
中国 [Zhong1 guo2] /China/
人 [ren2] /man/person/people/CL:個|个[ge4],位[wei4]/

and therefore the same pinyin romanization as the one proposed by the google translate gui tool (

https://translate.google.com/#view=home&op=translate&sl=zh-CN&tl=en&text=我是中国人 ):

Wǒ shì zhōngguó rén

But there would be problems for hanzi with several pronunciations as well as for cases where the accent depends on the context:

不好 bù hǎo
不是 bú shì

However, trying these cases with google translate, the same problems can be observed:

https://translate.google.com/#view=home&op=translate&sl=zh-CN&tl=en&text=不好 => Bù hǎo

https://translate.google.com/#view=home&op=translate&sl=zh-CN&tl=en&text=不是 => Bùshì (wrong accent)

So maybe the google translate gui tool is actually using a similar dictionary based approach and neglecting the context as well?

Regards,
Dietrich

Message last modified on Feb 15, 2020 08:21PM

di...@newskylabs.net <di...@newskylabs.net> #12Feb 15, 2020 08:40PM

Yet another comment:

Listening to the audio file which is returned by the google translate GUI as well, the accents are different :)

I suppose a deep learning approach is used over a corpus with both, written and recorded examples, so that the context dependency is learned as well.

Push the [listen] buttons on the following pages to verify:

https://translate.google.com/#view=home&op=translate&sl=zh-CN&tl=en&text=不好

https://translate.google.com/#view=home&op=translate&sl=zh-CN&tl=en&text=不是

Cheers, Dietrich

de...@gmail.com <de...@gmail.com> #13Apr 11, 2020 10:04PM

Very interesting, Dietrich. Google should just return pinyin together with the characters.

I scraped my way out for now.

ow...@gmail.com <ow...@gmail.com> #14Apr 12, 2020 04:35PM

Also registering desire for this feature.

mi...@google.com <mi...@google.com> #15Jun 17, 2020 07:56AM

Hello,

It seems that there exists a deprecated API (Transliterate API [1]) that does this exact thing. Given a specific text it transliterates it to a different script. This is a feature that would be nice if it could be added in Translation API.

[1] - https://developers.google.com/transliterate

so...@gmail.com <so...@gmail.com> #16Aug 12, 2020 05:55AM

+1 for this feature. Very much wanted.

ou...@gmail.com <ou...@gmail.com> #17Apr 1, 2022 09:05PM

Comment has been deleted.

Message last modified on Apr 1, 2022 09:05PM

el...@gmail.com <el...@gmail.com> #18Apr 27, 2022 04:58PM

+1 for this feature very much wanted and needed. In the meantime - am curious if there is another service (AWS or AZURE) that folks are using?

uj...@gmail.com <uj...@gmail.com> #19Aug 17, 2022 11:20AM

+1
@google, please bring this feature as early as possible

fb...@googlemail.com <fb...@googlemail.com> #20Oct 25, 2022 08:56PM

+1 for this!

ph...@gmail.com <ph...@gmail.com> #21Nov 28, 2022 01:23PM

+1 for this feature, would be very cool

se...@gmail.com <se...@gmail.com> #22Dec 12, 2022 02:09PM

pa...@gmail.com <pa...@gmail.com> #23Feb 9, 2023 01:31PM

How is this not a feature? Chat GPT can do it. :P

va...@gmail.com <va...@gmail.com> #24Feb 10, 2023 08:32AM

Is this even coming? Even I am waiting for this to be added in Google Translation.

yn...@gmail.com <yn...@gmail.com> #25Feb 14, 2023 01:31PM

waiting too..

ch...@gmail.com <ch...@gmail.com> #26Mar 24, 2023 08:45PM

I would like this as well

oz...@gmail.com <oz...@gmail.com> #27Mar 25, 2023 06:09AM

Once again Google has been caught sleeping and now their competitors have a better product. In this case, Chat GPT-3 gives me good results with prompts like this:

Consider this sentence in Traditional Chinese: 我喜歡吃鳳梨。

Annotate each character with its pronunciation in 注音. Present the results by rewriting the original sentence and having the pronunciation in square braces after each character.

The following characters are punctuation:
。，、
They should always be followed by a pair of square braces with a single empty space inside and nothing else.

Here's the output:

我 [ㄨㄛˇ] 喜 [ㄒㄧˇ] 歡 [ㄏㄨㄢ ˇ] 吃 [ㄔ ˄] 鳳 [ㄈㄥˋ] 梨 [ㄌㄧˊ] 。[ ]

Could use a few adjustments, but it's pretty nice! GPT-4 should be even better.

ma...@gmail.com <ma...@gmail.com> #28Dec 28, 2024 02:35PM

+1 pleaseeeeeee this would be very helpful