Faster translation ideas...please share your thoughts

Hi there,
I was just translating few pics (done a few now) and realized that about 80% of the "translation_request" tag stuff (which means roughly 3000 pics or so) no one cares if it gets translated.
Most things are easy Ah~ s and Oh~ s and some Oniichan daisuki~ stuff.
I mean...seriously: The request_translation tag is totally overused. It's not necessary on most pics and somehow even if it is, hardly anyone cares.
Now I want to make people happy with my (lets say mediocre) translations (I do it for fun so don't blame me). But for me, it's nearly impossible to pinpoint the pics folks want to get translated.
Three sentences above, I said that "hardly anyone" cares about translation of (lets say non-word-intensive) pics. In some cases these people want a tl so badly, they just comment something like: "Do want translation" or any other grammatically incorrect translation demand. Actually...this is what I and probably other translators are looking for: A sign that someone cares about a pic and a somewhat "real" translation request.

Long story short, I propose few ideas:

1. Metatag "Comment":
A few know that there is the metatag feature in the searchbar. Please refer to the Cheatsheet (http://danbooru.donmai.us/help/cheatsheet).
You can order results by e.g. score. Now in translation_request picture pool, it makes not much sense as the more higher the score gets the less interesting text there is (in lets say about 90% of all cases).

More interesting would be a order:comments metatag to see if a pic got comments (and therefore a chance of someone demanding a translation).

Example:

SEARCH:
Foo_Bar order:comments

Pic1 (/w 10 comments) Pic2 (/w 7 comments) etc etc.

I don't know how hard this is to code, but I guess it's more/less easily doable.

2. Translation request button:
The idea would be just a button or a textlink which every user can press one time per pic. It adds a vote to a translation_request counter.
Of course the counter should be viewable and preferrably searchable for us.
This would be a great improvement (I think), but on the other hand would probably mean a great deal of extra data (one extra row in the db for every pic).

It would look something like this:
SEARCH:
Foo_Bar order:requests

Result:

Pic1 (30 requests) Pic2 (25 requests) etc etc.

Hope you get the idea.

Phew. That was a long text. But it bothered me for quite some time. Hope this will stay up so some can see this and share their opinion.

So...please discuss.

In short, I like both ideas, although the button idea would probably still be subject to abuse.

I only add a translation_request tag if the image is impossible to understand without a translation, and I really want to know what it says.

You're right, a lot of people seem to add it to any old image with so much as an あ～. Perhaps it would be a good idea for someone to go through the tag and remove all the frivolous entries. I might actually do that next weekend or whenever I have a bit of time.

(As an aside, I also sometimes just avoid uploading stuff with lots of Japanese in it, since there's more than enough to translate already and I don't want to add stuff that's potentially useless. Since my Japanese is extremely limited I usually don't know if something like a 4koma is funny or not until it's translated.)

Personally, I try to translate everything that has a translation_request tag and looks like it might be within my ability (though I normally avoid explicit pics with extreme prejudice – they are, not to mention, a major supplier of the あ～ and はぁ we see so much). I don't really care about the amount of text aside from that.

Also, I usually spare little thought for how urgently someone wants to see something translated. I'm doing this mostly for my own entertainment/education after all, and besides, the image I'm translating is bound to turn up in someone's search eventually.

Still, I'll bet the order of precedence does matter to the less Japanese proficient, so it probably is worth looking into. I like the voting idea, however I'm afraid I couldn't comment on the technical implications.

surasshu said:
(As an aside, I also sometimes just avoid uploading stuff with lots of Japanese in it, since there's more than enough to translate already and I don't want to add stuff that's potentially useless. Since my Japanese is extremely limited I usually don't know if something like a 4koma is funny or not until it's translated.)

If it looks good, upload it anyway. Don't take the number of pending requests into consideration.

LaC said:
If it looks good, upload it anyway. Don't take the number of pending requests into consideration.

Well yeah, that's why I've uploaded for example riyo's stuff at all. I'm glad I did for some, but for others, after the translation I still don't get it. If it's really unfunny however, I feel like I'm wasting everybody's time, so I only upload if I'm fairly confident it's worthwhile on some level.

surasshu said:
If it's really unfunny however, I feel like I'm wasting everybody's time, so I only upload if I'm fairly confident it's worthwhile on some level.

The problem I see is rather the quite exquisite humor of the japanese 2chan people. Since I don't visit 2chan I can't say for sure, but I watch niconico-douga everyday. While some "Neta"-Videos are simply stupid, a lot of them have references no non-japanese would normally get. When the video is about 2chan I completely surrender to understand.
I guess that's the same for some 4-komas. Nevertheless I would be really glad if some get upladed, since I normally never stumble upon these comics.

Frevel said:
The problem I see is rather the quite exquisite humor of the japanese 2chan people. Since I don't visit 2chan I can't say for sure, but I watch niconico-douga everyday. While some "Neta"-Videos are simply stupid, a lot of them have references no non-japanese would normally get. When the video is about 2chan I completely surrender to understand.

I believe this can be said for 4chan as well. I've never been to 4ch, but after seeing the types of memes it spawns, I have no desire to ever go or to even try to understand their humor.

I think part of the problem is that you have to know at least a smidge of Japanese, sometimes, to know that some of those pics don't need a translation. I've seen "Oh man needs trans so bad, I have to know what they're saying" type comments on images that don't have a single complete Japanese sentence. People don't know what if anything they're missing, so they put the tag on it anyway.

As for a trans request voting mechanism, that sounds like a good idea.

I, for one, really appreciate your translation endeavors. (^-^)

I like both of the OP's ideas. Translation requests can't be abused if they're handled the same as votes, by the way.

Frevel said:
the quite exquisite humor of the japanese 2chan people.

... did you just call 2ch/2chan humor "exquisite"?

Well, that statement was smothered in sarcasm.

First of all thanks for your replies.
I guess a lot of people agree that there is indeed a fix needed.
Therefore my question:
Is there any administrative person here who can comment on the technical possiblity of my ideas? Or anyone who can point me to a sysop so I can show him this thread?

Well, albert reads the forum pretty regularly, so he's probably going to notice this thread at some point. However, to speed things up one thing you can do is submit a Ticket at http://trac.donmai.us, describing your solution and linking to this forum thread.

Totally agree with OP, it's almost impossible to find a picture really worth and demanding translation in the current state of things. I personally translate just random images that caught my attention but no way I can tell if these translations even noticed by anyone.

Ordering by comment count is not really feasible without adding an extra field, and I don't really want to add this when it's of interest to only a handful of people. Furthermore, I'm not convinced that comment count is a reliable indicator of complex text. People who don't know any Japanese will assume あー is some sort of word and will ask for a translation. I predict the comment counts will be in the 1-3 range and won't really give much data.

I would prefer a tag solution but of course this is worthless unless everyone knows. Maybe a pool, or a forum thread with translation requests.

As for a Request Translation link I'm kind of hesitant to do this because:

1) The interface is complicated enough already and you're asking casual users (who can't be bothered to add a tag) to find this small link when many people can't even find the Upload or Edit link
2) By making it easier to add the translation_request tag, you're also making it easier for images with trivial translations to get tagged as such

A pool might be a good idea actually... "Desperately Needs Translation" pool, or something. It would need to be heavily monitored however (of course).

A pool could at least sort the images by the time they were added to the pool.

Thank you albert for replying here.

Ordering by comment count is not really feasible without adding an extra field, and I don't really want to add this when it's of interest to only a handful of people.

I already thought this might be tricky/difficult to implement. As no one else can judge this matter better than you, it is clear that we shouldn't pursue this way anymore (aka count votes etc.). Perhaps its something to implement into the next major Danbooru version (Danbooru 2.0)?

Furthermore, I'm not convinced that comment count is a reliable indicator of complex text. People who don't know any Japanese will assume あー is some sort of word and will ask for a translation. I predict the comment counts will be in the 1-3 range and won't really give much data.

I agree that the comment count isn't indication enough, but since the comment count is already saved in the database (please correct me if I am wrong) such a feature wouldn't be much of a problem. Since it is "only" adding a metatag order:comments (with or without any parameters), no additional space is needed on the main interface.
The comment ordering would be useful for other things too, like finding good parody pics (no one would fav them anyhow..). I know it's far fetched but this metatag could mean a lot of fun :)

I would prefer a tag solution but of course this is worthless unless everyone knows. Maybe a pool, or a forum thread with translation requests.

The tag solution would probably be abused. Seriously, it would be the same as now (just with the new tags...).
A pool would be the easiest idea. I don't know if it is really user (new user) friendly though. Perhaps a text link below the "Add to pool" with the direct addition of the pic into the pool would make things easier. On the other hand the abusability would be much higher.

I disagree on the forum part: A lot are too lazy, some are too eager and in the end it's just gonna be a big list of links. I don't see a improvement there unfortunately :(

As for a Request Translation link I'm kind of hesitant to do this because:

1) The interface is complicated enough already and you're asking casual users (who can't be bothered to add a tag) to find this small link when many people can't even find the Upload or Edit link

I don't think adding one line to the "Add..." list will do any harm. The people who can't find the upload/edit link most likely don't add the translation_request tag; So these ones we can pull out of the equation entirely, since they aren't contributing anyway (but remember: a simple link is easier than a tag edit -> could mean more contribution from newcomers.)
Optional would be to set the color of the translation request textlink (temporarily) to red as an indication of something new.

2) By making it easier to add the translation_request tag, you're also making it easier for images with trivial translations to get tagged as such

Actually my idea was to connect the "voting system" with the text-link. Since that is technically improbable to implement without any problems (and is ofc timeconsuming), the idea moved to a textlink with the function to move that pic into a pool called "TL Request" (whatever). I never wanted to make the translation_request tag easier to use (we already have thousands of them anyway...).

The problem of abusing is of course the same. That's why I propose to test this "feature" (pool + textlink in color) for about two weeks and see what happens. Since it's only a minor programming (only the textlink) this shouldn't make any problems.

Any further comments and ideas are very welcommed. (Sorry for any bigger mistakes, I'm too lazy and tired to review my post for now)

P.S: Short question about the abusability: I'm interested if that is really a problem... Since most people watch the pics, only a slice will request a translation anyway. We can start implementing the pool idea and see if it is working without any interaction on mods' side.

I agree that a pool or separate tag would be helpful...there are multiple images that I've checked back on hoping to find it translated, but there's just so many.

I think the main problem is people just throw the "translation_request" on images just because it has japanese in it. So if there was some other way to differentiate between those it would be great.

All right, just a quick post about the "Comment"-Metatag possibility:
I looked at the database code and realized, that a "order by comments" isn't as easy as it sounds. I hope I'm wrong but perhaps a more technically adapt person can confirm it.
Just a quick (hopefully right) description:
All metatags regarding posts are just displaying db entries by rows. Here is the db-organization of the posts:

289 CREATE TABLE posts (
290 id integer NOT NULL,
291 created_at timestamp without time zone NOT NULL,
292 user_id integer,
293 score integer DEFAULT 0 NOT NULL,
294 source text NOT NULL,
295 md5 text NOT NULL,
296 last_commented_at timestamp without time zone,
297 rating character(1) DEFAULT 'q'::bpchar NOT NULL,
298 width integer,
299 height integer,
300 is_warehoused boolean DEFAULT false NOT NULL,
301 last_voter_ip text,
302 ip_addr text NOT NULL,
303 cached_tags text DEFAULT ''::text NOT NULL,
304 is_note_locked boolean DEFAULT false NOT NULL,
305 fav_count integer DEFAULT 0 NOT NULL,
306 file_ext text DEFAULT ''::text NOT NULL,
307 last_noted_at timestamp without time zone,
308 is_rating_locked boolean DEFAULT false NOT NULL
309 );

As most of you see, some rows are used for ordering (like # of faves, created at, height, width, score etc.) while others aren't meant for that (like IP adresses stored for voting purposes, etc).
The problem here is that the number of replies is not stored (directly) in this part. So a "display posts with >3 replies" isn't that easy.

Let's take a look at the comments-table:

119 CREATE TABLE comments (
120 id integer NOT NULL,
121 created_at timestamp without time zone NOT NULL,
122 post_id integer NOT NULL,
123 user_id integer,
124 body text NOT NULL,
125 ip_addr text NOT NULL,
126 signal_level smallint DEFAULT 1 NOT NULL
127 );

Here we can see the "post-id" row, which is probably used to indicate to which post someone replied to. This is (again: I'm not very adapt at that, please correct if wrong) the only possible way to search for any posts with a comments count higher than 1 (one is possible due to the "last_commented_at" row in posts) and this would require much more db-inquiries to do. One have to sort "comments" by post_id and then count all occurances (which sounds rather power/timeconsuming).