Videos on how google translate works

[Dawei3]

I’m a huge fan of google translate.  Is it prefect?  Of course not.  Is it an invaluable teaching tool and general translation tool?  Absolutely.  (for the right price too!)

 

Although the videos I’m offering aren’t new, I just found them.  I’m sharing them for anyone who has an interest in how google translate works and its limitations.

 

This older one gives the foundation for google translate (The book "Is that a fish in your ear" gives more perspective on this as well):

https://www.youtube.com/watch?v=_GdSC1Z1Kzs&ab_channel=Google

In summary, the video notes google doesn't have multilingual elves running app; that all of the translations are computer based, either by finding what humans have already translated or using the algorithms to translate.

 

If want to see the impact of neural networks and want to know more details on how they help google translate, check out (this is much more detailed):

https://www.youtube.com/watch?v=AIpXjFwVdIE&ab_channel=CSDojoCommunity

 

For those looking for fun & who like the song “La bamba”, I highly recommend the following on Lamba versus google translate.  I can watch it again & again looking to see which language is in each frame.

https://www.youtube.com/watch?v=06olHmcJjS0&ab_channel=Google

 

These videos will give you more perspective on machine translation.  I think it's cool.  

[phoneticsem]

I am also huge fan and student of a google translator. 

 

It perfectly translated this. 

为了了解新加坡人对本地传统 文化的想法,《新州日报》采访了 100名民众,发现全部都认为应该保 留本地传统文化。

选择保留传统美食的受访民众 认为失去传统美食对新加坡人而言 是很可惜的。本地的烹饪学府应该 教导学生制作本地菜肴,才能把传统美味延续 下去。政府承诺会在接下来几年内探讨如何为 年轻小贩创造有利的工作环境。

 

保留传统文化

手工艺品 30%                                                

裁缝 10%
杂货店 15%
传统美食 45%

 

1. 这项调查的目的是什么?

(A) 明白新加坡人对杂货店的看法。

(B) 教导学生要如何保留传统文化。

(C) 证明学生会想办法学习传统文化。

(D) 了解国人对本地传统文化的想法。

 

2. 根据调查,选择传统美食的民众占了

(A) 一小部分。

(B) 绝大多数。

(C) 刚好一半。

(D) 大约四分之一。

 

In order to understand Singaporeans' thoughts on local traditional culture, "New State Daily" interviewed 100 people and found that all of them believed that local traditional culture should be preserved.

Respondents who chose to keep traditional food felt that it would be a pity for Singaporeans to lose traditional food.  Local culinary schools should teach students how to make local dishes in order to continue the traditional delicacy.  The government has committed to exploring ways to create a favourable working environment for young hawkers over the next few years.

 

preserve traditional culture

 Handicrafts 30%

Tailors 10%
Grocery store 15%
Traditional Cuisine 45%

 

 1. What is the purpose of this survey?

 (A) Understand how Singaporeans view grocery stores.

 (B) Teach students how to preserve traditional culture.

 (C) Demonstrate that students will find ways to learn about traditional culture.  

(D) To understand Chinese people's thoughts on local traditional culture.

 

2. According to the survey, people who choose traditional food account for

 (A) A small portion.

 (B) the vast majority.

 (C) Exactly half.

 (D) About a quarter.  ,

 

 

[phoneticsem]

I want to understand, how 对本 is used in the context. 

为了了解新加坡人对本地传统 文化的想法,《新州日报》采访了 100名民众,发现全部都认为应该保 留本地传统文化

 

any one can share more on that?

[Jim]

On 11/13/2022 at 10:04 AM, phoneticsem said:

any one can share more on that?

It's 本地 not 本 on its own. The 对 addresses the whole subsequent noun phrase (本地传统 文化), something like "about" because it's Singaporeans' thoughts about...

[Dawei3]

On 11/12/2022 at 9:00 PM, phoneticsem said:

I am also huge fan and student of a google translator. 

 

It perfectly translated this. 

Its perfect translation is likely because google found a perfect translation.  I remember having google translate translate several very technical paragraphs about a drug from Spanish to English.  The translation was perfect.  Then I realized, it likely found the document translated (it was too perfect to be machine translations).  

 

The essence of how google translate works is similar to how google search works;  its view is if you're asking for a translation, likely someone else has already.  If they have, google translate provides it to you.  Google translate constantly searches the web for credible translations.  

 

When facing unusual translations, such as Icelandic to Farsi, google translate is less effective because there are fewer texts translated between these languages.  It may need to resort to seeing if books such Sherlock Holmes are available in both languages and use texts from the book (just as an example).

 

 

[Lu]

On 11/16/2022 at 4:43 AM, Dawei3 said:

When facing unusual translations, such as Icelandic to Farsi, google translate is less effective because there are fewer texts translated between these languages.

To my knowledge, and unless this has changed, Google translates everything through English. So it won't even attempt Icelandic to Farsi, it goes Icelandic -> English -> Farsi. The more different and un-European the language, the worse the results.

[Dawei3]

On 11/16/2022 at 4:12 PM, Lu said:

To my knowledge, and unless this has changed, Google translates everything through English

The book on translation, "Is that a fish in your ear", from 2011 discussed the basis of Google translate.  It scours the internet for high quality translations, such as those in the EU, for translations between uncommon languages, it uses whatever it translated works it can find.  It doesn't use English as an intermediate step.  It notes that books like Harry Potter, John Grisham's & others are used as the basis for translation.  It jokes that the real wizardry of Harry Potter may be its hidden power to support translation from Chinese into Hebrew. 

 

Also, it notes google translate is based on a mathematical framework developed by IBM in the 1980s (that had nothing to do with translation).  Direct translation from one language to another with this is more accurate than moving thru a 3rd language.  The videos give some perspective on this.  

 

[roddy]

While I haven’t done an in-depth comparison, DeepL seems to be doing very well. 
 

I’m increasingly sceptical of scepticism about machine translation. It’s very easy to trip it up by giving it something difficult, but more interesting is the increasing scope of material it can handle well. 

[Moshen]

Quote

but more interesting is the increasing scope of material it can handle well. 

 

So it's all the more interesting when you plunk in something that Google Translate can't handle intelligibly at all.  This happens to me frequently.  I'll stay on the lookout for the next time it happens and report here.

[Lu]

On 11/17/2022 at 2:55 PM, roddy said:

While I haven’t done an in-depth comparison, DeepL seems to be doing very well.

I keep hearing this about DeepL, but every time I try it, it disappoints. To be fair, I only try it when I come across something difficult.

[Zeppa]

DeepL is great. I use DeepL Pro to see how it handles my German-English legal translations, but I only find it helpful at suggesting the more general vocabulary in them. The technical legal vocabulary I have to be careful with.  I am more or less retired, however.

 

Translation memory programs nowadays usually let you (for an extra payment) import DeepL Pro versions into your current translation. It is widely used by human translators. The Pro version is private and allegedly does not use your text elsewhere, so your client's texts remain private. Its translation quality is no different from that of the free version of DeepL.

 

I'd just like to say, in case it's of any interest, that one problem of DeepL in professional translations is that it can be inconsistent, translating a term in more than one different way even if that term is the same in the original text. It also does not contain the translations of statute names, courts and so on that are recommended by the German authorities - some clients want those to be used. Of course, it sometimes makes mistakes, but it's great for many purposes.

 

Another advantage of using a human legal translator, although this is not what you are discussing, is that I can spot errors in the original text!