Skip to content

[FA] CH1 / P1-2#142

Merged
lewtun merged 10 commits into
huggingface:mainfrom
schoobani:choobani-contr
May 16, 2022
Merged

[FA] CH1 / P1-2#142
lewtun merged 10 commits into
huggingface:mainfrom
schoobani:choobani-contr

Conversation

@schoobani

@schoobani schoobani commented Apr 23, 2022

Copy link
Copy Markdown
Contributor

@jowharshamshiri and other dear collaborators, please provide me with you feedbacks on this PR.

Part 1:


Refer to the glossary on the use of the huggingface emoji.


Library names are not brand names, nor are they concepts, so no need to transliterate. They are shorter namespaces so just use the English form. For example we will transliterate 'Tokenizer' when it is used to indicate a concept but not when it is the name of a library.

( I used Fine-tune)
I don't think we should transliterate the word Fine-tune.

(I used دمو)
I don't think we should transliterate the word Demo.


also for lists separately. They render in LTR with the current version of Huggingface docs if you don't do this and the preview will look ugly. Refer to our previously submitted pages.

(I have translated AGI as هوش جامع مصنوعی, this is what persian wikipedia also refers to)
I defer to our new collaborators on whether we need to add context info about AGI. That info is always a google search away and the hard and fast rule is to not include ANY new contextual info. Refer to point no. 9 in the notes on translating at the bottom of the glossary. I'll soon be moving those out to a text file, per your suggestion.

(I can't really get what he means by robot immortality, maybe I can use حضور جاودانه روبات‌ها در میان انسان‌ها)
I think when the text talks about "robot immortality" it doesn't mean that robots will become immortal--they already are. It's talking about humans and machines merging and humans achieving immortality through that. Also does scale up translate to the Persian "افزایش مقیاس"? Again I defer to our new collaborators.


'اولین گام‌های آغازین' is redundant.


A couple of uses of 'Yek' as a substitute for English indefinite articles.


I don't think 'to democratize' translates well to the Persian "دموکراتیک کردن". The Persian has solely political connotations to me, but its use in the English text is not political at all. I defer to our new collaborators on a better equivalent.

(changed it to نویسند‌‌ه‌ی مشترک)
Do we have هم-نویسنده‌ in Persian?


I think your choice of تابع is better than the glossary's. Please include that update to the glossary in your PR.

(I also prefer using transliterated versions, so I changed them.)
I defer to our new collaborators on whether we should transliterate encoder and decoder or use the Persian equivalents رمزنگار and رمزگشا. Do we know what is used in Persian textbooks? and do we like that?

Part 2

(changed them all to پردازش زبان طبیعی)
Again hard-and-fast rule is not to expand the acronym NLP when the original text hasn't. There is some gray area here, as in the case of ML that we talk about in the glossary(Acronyms used for convenience in English). With some nuance I think expanding it into full Persian form is acceptable in some places in Page 2. If you choose to do that, please choose only one form, no need to include the acronym in ellipses.

(In this case couldn't think of a better substitute)
Persian sentences follow the format Subject-Object-Verb. That means it is generally expected that the sentences end in a verb. Again there is a lot of gray area here and we need to apply nuance. But try to keep that in mind. For example maybe don't end sentences in "هستند یا نه". I say maybe bc if you do, that is your creative choice and I respect that.

(I prefer دسته‌بندی and I think the glossary says دسته‌بندی too)
I propose the use of "تشخیص نوع" in place of "دسته‌بندی" in Page 2, the verb 'to classify' is not used as a technical term there


Is "بیان کردن" our best equivalent for 'represent'?


Chapter is فصل and section is بخش. The glossary is your friend. I'm sorry I had this in my local fork but have not finished Page 2 of Chapter 2 to submit a PR with a glossary update yet. I thought it was already in there. Forgive my assumption.

@jowharshamshiri

Copy link
Copy Markdown
Contributor

Hi @schoobani jan, thanks for responding to these points separately. I'll get to this tonight. ❤️❤️

@jowharshamshiri jowharshamshiri left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Great job ❤️. Please include additions to G.

Comment thread chapters/fa/chapter1/1.mdx Outdated
<div dir="rtl">
# مقدمه

## به دوره‌ی آموزشی هاگینگ‌فِیس خوش آمدید

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

T/point1: we deviate from the glossary here and use 'ء' instead of 'ی' as a possessive suffix. This should be "دوره آموزشی"
Will call this Issue 1 for the rest of this review.

Comment thread chapters/fa/chapter1/1.mdx Outdated

## در این دوره چه چیزهایی را می‌آموزیم؟

دید کلی کوتاه از مباحث این دوره‌ی آموزشی:

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Same as above.

Comment thread chapters/fa/chapter1/1.mdx Outdated
دید کلی کوتاه از مباحث این دوره‌ی آموزشی:

<div class="flex justify-center">
<img class="block dark:hidden" src="https://huggingface.co/datasets/huggingface-course/documentation-images/resolve/main/en/chapter1/summary.svg" alt="دید کلی کوتاه از مباحث این دوره‌ی آموزشی">

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for changing the alt on these links. I don't know if I did this on my pages, will check to fix this there if I missed it.

Comment thread chapters/fa/chapter1/1.mdx Outdated
</div>


- از فصل ۱ تا ۴ مقدمه‌ای از مباحث‌ پایه‌‌ای کتابخانه‌ی ترنسفورمرز هاگینگ‌فِیس ارائه می‌شود. در پایان این فصل، شما با شیوه‌ی عملکرد مدل‌های ترنسفومر آشنا می‌شوید و می‌آموزید که چگونه از یک مدل در [هاب هاگینگ‌فِیس](https://huggingface.co/models) استفاده کنید، آن را برای مجموعه داده خود fine-tune کنید و نتایج خود را در هاب به اشتراک بگذارید.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I propose کوک‌ کنیم instead of fine-tune per @hamedonline 's suggestion. Whatever we do let's not transliterate.

Comment thread chapters/fa/chapter1/1.mdx Outdated
به دوره‌ی آموزشی هاگینگ‌فیس خوش آمدید!
<Youtube id="00GKzGyWFEs" />

در این دوره‌ی آموزشی، پردازش زبان طبیعی را با استفاده از کتابخانه‌های اکوسیستم [هاگینگ‌فِیس](https://huggingface.co/) یعنی [Transformers](https://github.com/huggingface/transformers), [Dataset](https://github.com/huggingface/datasets), [Tokenizer](https://github.com/huggingface/tokenizers), [Accelerate](https://github.com/huggingface/accelerate) و همچنین [هاب هاگینگ‌فِیس](https://huggingface.co/models) می‌آموزید. این دوره کاملا رایگان و بدون تبلیغات است.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The names of Datasets and Tokenizers is missing an s at the end.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Please add footnote to introduce the acronym NLP.

Comment thread chapters/fa/chapter1/2.mdx Outdated

با این حال پردازش زبان طبیعی صرفا به متون نوشتاری محدود نمی‌شود و برای چالش‌های پیچیده‌ی بسیاری در مسائل تشخیص گفتار و بینایی ماشین راه‌حل ارائه می‌کند. برای نمونه می‌توان از تولید متن از یک فایل صوتی و یا تشریح یک تصویر، نام برد.

## چرا چالش‌برانگیز است؟

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I propose "چرا این مبحث چالش برانگیز است".

Comment thread chapters/fa/chapter1/2.mdx Outdated

## چرا چالش‌برانگیز است؟

کامپیوترها اطلاعات را مانند انسان پردازش نمی‌کنند. برای مثال زمانی که ما جمله‌ای مانند من گرسنه هستم را می‌خوانیم، به سادگی معنای آن را متوجه می‌شویم. همچنین زمانی که دو جمله‌ مانند من گرسنه هستم و من ناراحت هستم را می‌خوانیم، بسادگی می‌توانیم تشخیص دهیم که به چه میزان این دو جمله با یکدیگر تشابه دارند. برای مدل‌های یادگیری ماشین، چنین مسائلی به مراتب سخت‌تر است. متن باید به ‌شیوه‌ای پردازش شود که به مدل امکان یادگیری از آن را بدهد. و با توجه به اینکه زبان پیچیده است، باید در پیاده‌سازی این مدل‌ها بسیار دقت کنیم. تحقیقات بسیاری انجام شده است تا نشان دهند چگونه می‌توان متن را توسط کامپیوتر بیان کرد. در فصل بعدی به برخی از این شیوه‌ها نگاهی میاندازیم. No newline at end of file

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I propose "چگونه می‌توان متن در کامپیوترها مدل کرد" instead of "چگونه می‌توان متن را توسط کامپیوتر بیان کرد".

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I added the missing را

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I added the missing را

I saw represent on a page in chapter 2 and actually بیان کردن worked really well there, so thank you for this word. I still think when I read it here, it sounds to me like "How do we get the computer to speak the text" but defer to you on this.

Comment thread chapters/fa/chapter1/2.mdx Outdated

## پردازش زبان طبیعی چیست؟

پردازش زبان طبیعی زیرشاخه‌ای از زبان‌شناسی و یادگیری ماشین است که تمرکز آن بر درک همه‌ی جوانب زبان انسان‌ها است. هدف مسائل صرفا درک کلمات بصورت مجزا نیست، بلکه جمله، متن و در مجموع‌ زمینه‌ای است که آن کلمه در آن به کار رفته است.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Issue 2.

Comment thread chapters/fa/chapter1/2.mdx Outdated

پردازش زبان طبیعی زیرشاخه‌ای از زبان‌شناسی و یادگیری ماشین است که تمرکز آن بر درک همه‌ی جوانب زبان انسان‌ها است. هدف مسائل صرفا درک کلمات بصورت مجزا نیست، بلکه جمله، متن و در مجموع‌ زمینه‌ای است که آن کلمه در آن به کار رفته است.

مسائل متداول پردازش زبان طبیعی بهمراه برخی مثال‌های آن را در این لیست می‌بینید:

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Issue 2.

Comment thread chapters/fa/chapter1/2.mdx Outdated
- **استخراج پاسخ از یک متن**: پاسخ به سوالات با استفاده از اطلاعاتی که در متن زمینه ارائه شده است.
- **تولید متن جدید از یک متن ارائه شده**: ترجمه‌ی متون به دیگر زبان‌ها، خلاصه‌سازی متون.

با این حال پردازش زبان طبیعی صرفا به متون نوشتاری محدود نمی‌شود و برای چالش‌های پیچیده‌ی بسیاری در مسائل تشخیص گفتار و بینایی ماشین راه‌حل ارائه می‌کند. برای نمونه می‌توان از تولید متن از یک فایل صوتی و یا تشریح یک تصویر، نام برد.

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Issue 2.


کامپیوترها اطلاعات را مانند انسان پردازش نمی‌کنند. برای مثال زمانی که ما جمله‌ای مانند من گرسنه هستم را می‌خوانیم، به سادگی معنای آن را متوجه می‌شویم. همچنین زمانی که دو جمله‌ مانند من گرسنه هستم و من ناراحت هستم را می‌خوانیم، بسادگی می‌توانیم تشخیص دهیم که به چه میزان این دو جمله با یکدیگر تشابه دارند. برای مدل‌های یادگیری ماشین، چنین مسائلی به مراتب سخت‌تر است. متن باید به ‌شیوه‌ای پردازش شود که به مدل امکان یادگیری از آن را بدهد. و با توجه به اینکه زبان پیچیده است، باید در پیاده‌سازی این مدل‌ها بسیار دقت کنیم. تحقیقات بسیاری انجام شده است تا نشان دهند چگونه می‌توان متن را در کامپیوترها مدل کرد. در فصل بعدی به برخی از این شیوه‌ها نگاهی میاندازیم.

[^1]: Natural Language Processing (NLP)

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Sorry if I was vague in my comments. I meant the first use on page one only where the source material explicitly has the acronym in ellipses and long story short we have footnotes for NLP on page one in your translation. On page 2 this is implicit by juxtaposition with the title for the next paragraph and no footnotes are required, I think. Same goes for the rest of the book, no footnotes for NLP.

<div dir="rtl">
# واژه‌نامه

| معادل در منبع | معادل در ترجمه |

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I get what you did here and want to thank you for the effort and the attention to detail Saeid jan. With the initial set of words I used an online service that automatically used a calculated fixed width in markup but then as we added new words and phrases we couldn't keep up that width, as will happen again in future with this version as well. I also have multiple diverging branches for pages half-done, that add lines to this file in arbitrary places that I don't really remember, and I'm sure other collaborators do as well. If we do this now, diffing becomes a manual task for us and one that is very time-consuming now that we are at 100+ words, as evidenced by the fact that right here I can't really tell what words you added, and neither can the Github diff. Please revert this to the old version and only add the new words. That version will greatly reduce the overhead for other collaborators when diffing.

That ofc holds true only for the beginning of the file to the end of the long word list. From then on please go ahead and apply what ever formatting you think is cool. In fact, while you're at it, if you want, remove the English footnote at the end. I think we are all on-board with our equivalent for inference now, and this line in English, no longer serves any purpose.

If the only word you added is "collaborative filtering" at the end, please capitalize.

@jowharshamshiri

Copy link
Copy Markdown
Contributor

@schoobani jan, it seems the refs for footnotes 2 and 3 on page one are switched. :)

courtesy of @kambizG jan.

@schoobani

Copy link
Copy Markdown
Contributor Author

@jowharshamshiri I can't resolve the conflicts. 🤔

@jowharshamshiri

jowharshamshiri commented May 15, 2022