Jichi's Annot Player -- allows annexing manual translations to ANY game (without hacking)

[Guest]

Update: Update on Jichi’s Annot Player.

See A Revolution for Fan Translation–it is now possible to sub any visual novel without hacking (by jichi)

Visit Annot Player.

Also Looking for a coder to help jichi with Annot Player.

(I'll populate this thread with more info soon)

[REtransInternational]

Good to see this initiative is gaining activity.

His 4th item is particularly interesting, as a SQL-based backend means many existing standards-based tools can be used to process the translation data, and it naturally fits with the formats we use for text extraction/insertion, making realisation of one-click patch creation/download considerably easier.

We've done some (very minimal) planning around the 2nd item; I'd be interested in hearing what sorts of ideas he has. Perhaps using manually input translation data for sMT engine training? There's been a lot of linguistics research into this but for general MT and not specifically targeted at VNs, so maybe a targeted approach may yield much better results.

I'd say definitely invite Jichi here if you haven't already. It looks like he'll be an invaluable asset.

[Guest]

Good to see this initiative is gaining activity.

His 4th item is particularly interesting, as a SQL-based backend means many existing standards-based tools can be used to process the translation data, and it naturally fits with the formats we use for text extraction/insertion, making realisation of one-click patch creation/download considerably easier.

We've done some (very minimal) planning around the 2nd item; I'd be interested in hearing what sorts of ideas he has. Perhaps using manually input translation data for sMT engine training? There's been a lot of linguistics research into this but for general MT and not specifically targeted at VNs, so maybe a targeted approach may yield much better results.

I'd say definitely invite Jichi here if you haven't already. It looks like he'll be an invaluable asset.

Not just a invaluable asset -- he's ultra polite as well!

[Tromend]

Well I see the initiative really good, but th only point it's been bugging me is: How does annot filter Trolls whi start putting false Translations or only garbage? I'm quite interested in knowing how to stop this kinds of attacks that are more than possible, and if there is no way to stop it right now, why don't we think of a way to stop it and consult it with Jichi?

[Guest]

How does annot filter Trolls whi start putting false Translations or only garbage? I'm quite interested in knowing how to stop this kinds of attacks that are more than possible, and if there is no way to stop it right now, why don't we think of a way to stop it and consult it with Jichi?

Well Jichi wrote, "he's working on an underlying user system for managing subtitle permissions."

I should put into a picture how I feel i'd like it to work, i'll do that tomorrow or soon

[REtransInternational]

Storage is cheap and text requires a minimal amount of it, so one easy solution is to store all submitted versions of a line, along with their associated user. Others can set a preference for what user's lines they'd like, and essentially give their own ranking to the translations. E.g. someone likes user1's translations and thinks user2's are ok, then he can set user1,user2 in that order and the system will display, for each line, submissions from user1, and if there aren't any, then user2's, and if there aren't any, then default to whoever else was the last to have submitted a translation for that line.

This will also solve any edit wars; if two translators disagree on how one line should be translated each user can choose and receive whose lines they prefer.

[Clephas]

How about registering as a fantranslator by proving you can translate a random set of complex lines taken from various vns/eroge infamous for their wordiness? You could then authorize them for various game types based on how well they managed it. Well, that would require a lot of personal management... so it might - probably - not be workable. Still, I had to throw it out there as an idea. Just letting any random guy register a random translation will result in some truly... unfortunate line choices.

[Guest]

By the way, Jichi is in the process of recoding almost the entire thing (from C++ -> Javascript) to be more user intuitive and so that it allows people to write extensions to it easier.

This is NOT a prototype image!! XD

This is just my proposal I mocked up in Photoshop:

He's going to follow some suggestions I made.

Here is the suggestion for Annot Player Lite: https://docs.google.com/drawings/d/1ur46wQjJlTCE6oHcVK_jUBr4-HOQMsLH2u4xm2546EI/ (you have to click the Zoom in Tool to Zoom in. very annoying)

The underlying idea is:

"it doesn't require much learning and so people are more willing to use it. If it requires learning then people just won't use it. "

[Guest]

[Tay]

I love the mockup! Beautiful, smooth and --best of all-- so logical that I don't think I'd need instructions to jump in and start!

[Guest]

I put an update post here:

https://visualnovelaer.wordpress.com/2012/11/05/update-on-jichis-annot-player/

annot player is going well, we could get a pre-alpha release next weekend

[jichi]

Well Jichi wrote, "he's working on an underlying user system for managing subtitle permissions."

I should put into a picture how I feel i'd like it to work, i'll do that tomorrow or soon

Regarding the anti-spam feature, I think a global shared filter like the one in nicovideo, or a statistics-based shared blocking system (so-called 云屏蔽) like the one in bilibili might help.

Regarding the underlying user permissions system, I don't have a whole picture of where it should go. Just let me know if you have suggestions on it.

[Steve]

Regarding the anti-spam feature, I think a global shared filter like the one in nicovideo, or a statistics-based shared blocking system (so-called 云屏蔽) like the one in bilibili might help.

Regarding the underlying user permissions system, I don't have a whole picture of where it should go. Just let me know if you have suggestions on it.

I was trying annot player for a game nobody subbed (futaba channel 2.5, I even added few TLs but then deleted it because I wasn't satisfied with my English wording) so I have some experience.

What I would recommend for the separate Visual Novel Reader spawned off annot player would be having a list of all people who have comments on the specific game and being able to turn off all of them except the one you choose the TL from. (so for example if I choose DC3, I would disable everyone's comments except Aaeru.

For the collab work, there might be feature to "link" specific contributors together and you wouldn't be able to turn them all off one by one, except all together as a "DC3 - randomgroup TL" for example, where the translators would be creating these links themselves, if they are translating a project with multiple people.

That way users should be able to get only the TL they want from the people/group they want etc.

Having possible teams within the system for new translation groups trying to use this new technology might also be a good idea, you would then see "DC3 random team" that has the translation done and they could invite people.

[jichi]

I was trying annot player for a game nobody subbed (futaba channel 2.5, I even added few TLs but then deleted it because I wasn't satisfied with my English wording) so I have some experience.

What I would recommend for the separate Visual Novel Reader spawned off annot player would be having a list of all people who have comments on the specific game and being able to turn off all of them except the one you choose the TL from. (so for example if I choose DC3, I would disable everyone's comments except Aaeru.

For the collab work, there might be feature to "link" specific contributors together and you wouldn't be able to turn them all off one by one, except all together as a "DC3 - randomgroup TL" for example, where the translators would be creating these links themselves, if they are translating a project with multiple people.

That way users should be able to get only the TL they want from the people/group they want etc.

Having possible teams within the system for new translation groups trying to use this new technology might also be a good idea, you would then see "DC3 random team" that has the translation done and they could invite people.

What do you think, Aaeru sensei?

[REtransInternational]

The editors of wikipedia does not get to choose whether his edits are locked permanently to his version, no matter how brilliantly he wrote it or how many thousands of hours he spent researching that article.

Good point. It is odd that TLwiki, despite being a wiki, went almost the exact opposite route from Wikipedia...

Being able to show multiple versions of a line might be interesting, if only to compare how different translations can be.

Regional/dialect issues also suggest it is good to give the user more choices.

The preference filter suggested above would also help, but should default to any-available --- that's the goal after all.

[Down]

But to date, there is no translation group that willingly translates a game unless someone else isn't already doing it.

Actually, there is, but it's more like the exception confirming the rule. Commie and Cokesakto (guy that translated the Kara no Kyoukai LN) are both working on Mahoutsukai no Yoru. It's not really 'competitive', Commie guys said that Cokesakto's TL of the demo sucked, but he doesn't seem to give a fuck.

Overall, I think you're right, but having only one translator doing the whole thing does help to keep the translation consistent and coherent, with someone doing edit/QC after. Crowdsourced translation is a great idea because there are people out there willing to participate but not having the time/resources to do the whole translation, but it does lead to TL inconsistencies and people randomly putting machine-translated and incoherent lines, thus slowing down the whole process. (Of course the Annot system helps dealing with those issues).

I do realize however that people doing the TL alone are mainly doing it for the e-peen.

[Guest]

imo Mahoyo is a good pick for competition, too high profile, and shorter than most others. But that's how I define competition though, "I know I can take a bigger slice of that rep for Mahoyo than you can". (with my TL quality/speed)

that's why i think the crowdsourced solution is better suited to moege and not the ones that are more convoluted.

the average VN is really low level language and the plots are too predictable, they're really hard to mess up.

[zoom909]

that's why i think the crowdsourced solution is better suited to moege and not the ones that are more convoluted.

the average VN is really low level language and the plots are too predictable, they're really hard to mess up.

The Dos and Don'ts of Quality Translations (draft version):

DO strike a balance, between keeping the original meaning of the Japanese, and rendering the English as naturally as you can.

DO translate with the aid of multiple dictionaries and thesauri.

DO maintain your finished work, providing updates whenever someone points out an error you think is legitimate

DON'T crowdsource (dividing routes among translators is OK, but individual lines? No.)

DON'T translate into any language other than your native.

DON'T translate to/from an intermediate language.

DON'T use MT (that should be obvious)

Even supposedly simple "moege" can be affected when you violate these guidelines.

Edited February 1, 2013 by zoom909

[zhurai]

I don't really think splitting every line and translating like that isn't a good idea as each translator has a different way of translating a line, that the final piece will have consistency problems.

[REtransInternational]

The Dos and Don'ts of "Serious" Translating (draft version):

DO strike a balance, between keeping the original meaning of the Japanese, and rendering the English as naturally as you can.

DO translate with the aid of multiple dictionaries and thesauri.

DON'T crowdsource (dividing routes among translators is OK, but individual lines? No.)

DON'T translate into any language other than your native.

DON'T translate to/from an intermediate language.

DON'T use MT (that should be obvious)

Even supposedly simple "moege" can be affected when you violate these rules. We've all seen the evidence, haven't we? Well, I guess rules are made to be broken...

That's been discussed at length here. Fuwanovel is about exploring new ways to produce translation more efficiently, and to break away from the "old culture". The old culture's attitude is "you must do it our way" and they embody the belief that quality is an absolute, sacrificing many other things that are often just as important. There'll always be tradeoffs, and we view quality/accuracy as not the only factor in delivering value to the consumer. Make it easy and enjoyable to increase participation, finding new processes to explore the tradeoffs, exploiting technology, that is what the new culture is about.

[zoom909]

OK, I read the other thread. Now that my initial surprise has worn off...

I admit that what you said, about there being tradeoffs, is true. So, let me change "rules" to "guidelines."

If you have a good reason, if you want to sacrifice some quality in order to obtain efficiency and so forth, I won't say you MUST follow them.

[Guest]

Hey I deleted one of my posts above.

reading over it, it really is quite horrible. I was supposed to leave a message in place of the deleted post, but for some reason I messed up and it didnt do that...

anyway sorry for that.