[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/jp/ - Otaku Culture

View post   
  [?]In ghost mode, only threads with non-archived posts will be shown
File: 16 KB, 500x400, 1267632838067.png [View same] [iqdb] [saucenao] [google]
4621444 No.4621444 [Reply] [Original]

Hello, /jp/.

I'm wondering if a utility like this exists:

It would go through an image or a series of images and detect japanese characters, convert them to text, run it through google translator or some other service and provide a quick and easy (maybe not as accurate, though) translation of an image, or perhaps a manga or doujinshi.

Would be useful, especially for untranslated images/works which won't be translated for a while.

>> No.4621450

They'd have to be pretty high quality scans.

>> No.4621449

interesting idea... would be nice to have a program like that

>> No.4621455
File: 84 KB, 571x800, Aka_no_moku_c.jpg [View same] [iqdb] [saucenao] [google]
4621455

All the pieces exist in the form of open software. Write some glue, and you're golden.

>> No.4621460

No one makes useful software.

>> No.4621466

learnin japanese

>> No.4621469
File: 54 KB, 427x600, Day_of_revolution_c.jpg [View same] [iqdb] [saucenao] [google]
4621469

>>4621455
>translation of an image, or perhaps a manga or doujinshi
>manga or doujinshi
Ok, that might not be automatable for free yet.

>> No.4621484

There are programs for the blind and such which read books for them by noticing characters on the scanned pages.

but they're asspensive.

>> No.4621494

There are plenty of OCR programs out there. Just write a script that will automatically copy the text to your clipboard (should be easy as hell) and have TA running.

>> No.4621551

http://twinbridge.com/detail.aspx?ID=342

translation program that uses OCR, I believe

>> No.4621562

if it was that easy we wouldnt have to go through people like desudesu to get our touhou porn translated.

>> No.4621570

free OCR program that does japanese http://www.4shared.com/get/103693432/f9a01a7f/SmartOCRlite107.html

>> No.4621580

>>4621570

>more like ocr program IN japanese

>> No.4621586

>>4621444

There is actually an Adobe program called "Adobe Acrobat" that does exactly what you're describing, but for english. So you'd scan a book, run it through the text recognition and have text that you could copy and paste into another document.

As far as I'm aware, the technology for it isn't all that advanced, I'm just not sure if there's a japanese version of it.

>> No.4621603
File: 170 KB, 1022x695, ohohohoho.jpg [View same] [iqdb] [saucenao] [google]
4621603

Something...like this?

>> No.4621604

http://maggie.ocrgrid.org/nhocr/

>> No.4621607

>>4621603

OHOHOHOHOLY FUCK

>> No.4621611

>>4621603

what is that program?

>> No.4621614

Its not the translating of the language thats the hard part.

Its figuring out the grammar and references.

>> No.4621615

>>4621603

GOOD GOD THIS WOULD MAKE THINGS SO MUCH EASIER.

Downloading instantly.

>> No.4621626

>>4621615

downloading from where?

>> No.4621633
File: 51 KB, 800x800, 8ae556221471f8d6f3e951394e6aaea2.png [View same] [iqdb] [saucenao] [google]
4621633

>>4621626

Yea, I noticed that after I posted.

Fucking image has no source.

>> No.4621639

>>4621603
Looks like at least 25% of the kanji are wrong in that screenshot.

>> No.4621653

>>4621639
I'll take that error rate over looking up almost all of them.

>> No.4621668

>>4621603
『読んde!!ココ

>> No.4621688

>>4621653
enjoy your poor translation, bucket dolphin washing microwave

>> No.4621689

>>4621668

I love you.

>> No.4621693

>>4621688
electronic brain pancake crystal elderly?

>> No.4621696

>>4621688
... of course I'd look up the ones it gets wrong, idiot.

>> No.4621704

>>4621639
lol, they sure are. Never noticed that.

>> No.4621703

>>4621688

Not necessarily. A quick google with most of the sentence in contact but the few kanji errors will most likely produce the result you are looking for.

>> No.4621710

Software like this has been around for ages. Its not as useful as you think.

>> No.4621712

Uh, this can't work (well) for several reasons:

1. OCR is shit, especially on handwritten english
2. OCR is even shittier for japanese because there are a million different characters that are all much more complex and/or similar to each other
3. Machine translation, especially Japanese machine translation, is shit. If it wasn't we'd all be playing VNs now instead of posting on /jp/.

>> No.4621719

>>4621712
Not to mention the number of false positives and skipped characters.

>> No.4621729

>>4621710
All I want is to not have to resort to radical search so often. That's it. Can it not even do that?

>> No.4621734

Easiest way to look up characters is w/ a DS dictionary.

>> No.4621743

>>4621729
You will have to anyway when you do your proof read of the garbage it dumps out. You will have to choose what character is what and there will be a lot of false positives. They also miss shit. There is a reason no one does it this way.

>> No.4623177
File: 3 KB, 104x126, 1268065905341.jpg [View same] [iqdb] [saucenao] [google]
4623177

>> No.4623193

You need to know at least some Japanese in order to use OCR, you're going to need to be constantly fucking correcting it. It's not meant to be your little personal translation station, it's a fucking tool meant to help in transcribing of texts. And much help it does not give since you still need to go through it anyway.

>>
Reply to thread [?]Don't expect anything heroic. Your post will not be uploaded to the original board.
Name
E-mail
Subject
Comment
Action