Brood War Korean Translations
This was an fun read, as someone who's both a Korean BW player and a speech recognition researcher.
It's interesting to note that the original Korean transcription already has many errors, seemingly (and impressively) corrected by LLMs later on. For example, 12 안마당 빌드 (12 courtyard build) is actually 12 앞마당 빌드 (12 frontyard build), which might have been more understandable to BW players. Similarly 투에처리 빌드 (processing-at-two build? makes no sense lol) should have been transcribed 투해처리 빌드 (two-Hatchery build).
Therefore it may also be helpful to directly feed the slang dictionary into Whisper's inference process using contextual biasing. There are lots of ways to do this, but the simplest would be to increase the probability of slang words in the dictionary in the final prediction layer of Whisper by a constant factor. This is fairly easy to implement, for example by using HuggingFace's library: https://huggingface.co/docs/transformers/en/internal/generat...
Don’t let the title fool you: this is anextremely thorough and creative take on translating and making more approachable the commentary of StarCraft.
As the author rightly points out, in its 27 years of existence, commentary around the game has become a domain specific language. Not just Korean or English.
This approach of automated scripting and using AI to understand roughly what was said and then make it coherent is really cool.
LOL, as a non-native English speaker, reading this reminds me of EXACTLY the same problem of translating many things, but more precisely, computer articles and software development.
There’s a huge amount of terms that are difficult to translate (sharding? Hash?). The only real solution is to adopt them to your language, more or less adapted, which is what happens over time. But it requires a community that, to some degree, is able to cross the gap between the languages. In this case, learning English.
Talking about software development in Spanish (my native language) is a succession of imported terms from English.
I don’t think there’s a good way of doing that, and I’m interested to see how automatic translations deal with it, because the only way this can work is with a process of mixing both language in a social way and see what terms evolve from that process.
And you need, in the terms the post describes, people that know Korean at least in a non-fluent way. And the game itself, of course.
Kinda funny that in an article about translations, the author gets signal-to-noise completely backwards. A high signal to noise (over 9000) is very good. It means you are getting a lot of signal with very little noise. Decreasing signal to noise means getting more noise.
Even when Google Translate got pretty good I was not really able to effectively translate Chinese or Japanese text about Go (the game). I had similar issues to the ones mentioned in this post. Many Chinese and Japanese words (e.g., "ko") have a very specific meaning in the context of Go, but they also have regular meanings (e.g., "robbery") in more normal contexts, so Google Translate would translate text in a generic way, which made everything unintelligible. With modern LLMs, I can now preface my translation requests with instructions such as "I am going to ask you to translate some Chinese text accompanying weiqi diagrams. Your translations should be idiomatic and not shy away from Go jargon. For example, 拆 = extension, 夹 = pincer, 刺 and 觑 = peep.", and it does a fantastic job, enough for me to basically read anything I want. It was lucky for me that evidently enough Go material already existed in the training set that I didn't have to do anything more special.
(Some chess corrections, in case the author is reading: the moves at the start of chess games are called openings in English, not openers; there are not distinct white-piece and black-piece openings, although of course an individual player will probably study a given opening from the point of view of one side or the other; their study is considered fundamental all the way up to the highest level, in fact more so as you increase in skill; and the Sicilian variation in question is the Najdorf, not Najdork.)
If the author sees this: with yt-dlp you can download lower quality versions of videos to save bandwidth, like so:
yt-dlp -f "bv[height<=720]" <url>
That will download up to 720p quality.
Dumb question from someone who only played money-maps as a kid:
What do the numbers in front of the building mean? 12 Hatcheries seems like… well, 12 seems like a possible but implausible number of hatcheries to build (hypothetically it is possible of course). And 12 spawning pools is obviously not useful. So that makes me think it is the position in the build order list. But, they list other builds, like:
> The second is the 12 Hatch, 12 Pool, 12 Gas
Which doesn’t make a ton of sense in with that parsing. I mean it must not be a straight list. Maybe it is a tree, and 12 is the depth for this building? But that seems late, I can’t think of 11 buildings to build before gas. Maybe they include units too? Or maybe just drones/overlords?
I loved this article, thanks for writing it.
I attempted playing a few world cyber game US regional matches and I was always amazed how much faster everyone else was. Then I remember when they live streamed it from Korea and I saw how fast they played and I was blown away. From a strategy point of view, something so basic about the game that I missed was when a blog introduced me to some math for a protoss zealot power up that defeated a zergling in 2 hits rather than 3. That's when I realized this is a chess game and I got hooked.
I get that it's "wrong" but I really like the translation of "natural expansion" to "courtyard"
I really wish someone with the resources and connections could get in touch with South Korean broadcasters in order to get access to their archives so that more historical games could get uploaded and re-commentated for a western audience.
My favorite Brood War slang term is Ee Han Timing [0]: basically when you take a risky build that has to do damage in a small timing window. A ton of exciting Brood War moments come from exploiting tiny timing windows.
How far off are we from local immidiate voice translation? Something on my computer that translates all spoken words by my computer, keeping tone and intonation?
I've always wanted translations of Korean game broadcasts (GSL/ASL). I'm curious what balance their commentators strike between game analysis and entertainment, and whether there's a disparity between Korean and Foreigner game analysis. Do they have a Tastosis?
This is great work, but really it seems that anyone who is a truly dedicated Brood War amateur should knuckle under and learn Korean. It really isn’t that unapproachable: with a decent approach (try https://refold.la ) you should be able to get somewhere useful in maybe “just” 3000 hours.
This is amazing. I'm getting chills!
As a side note, I have gotten into watching a "foreigner" BW channel everyday called ArtosisCasts. The videos are very strategic and high level commentaries on games as they are watched for the first time, with some after-match highlights for especially interesting maneuvers. It has really made me appreciate the depth of the game, as well as explain how I was so bad at it in high school. It has actually made me think a lot about startups, economic optimization, and how you approach the "meta" of any activity you're undertaking.
I've been watching BW out of Korea since 2007. Previously also played but it's been many years. This is really cool, thanks for sharing!
There are two YouTube channels I wanted to take the opportunity to shout out, the first one does English translations of Korean BW content, and the second one provides commentary on recent tournaments like the ASL with a little bit more depth then Tasteless and Artosis (no hate but to me their commentary is too off topic and they miss basic things about build orders).
for any of the lucky 10000, like me, who were left wanting to see what this game looked like:
https://youtu.be/Nm-PXmOELAw?si=Z-RXbdqNzkSF3cqx
my brief search didn't show me any more obscure Korean only strategy videos, so maybe this one is just for the lowly foreigners :(
Reminds me, many years ago someone paid me to translate a Korean wiki article about some League of Legends pro player to English. No idea why, most of it was random trivia, it didn't contain any notable insights. But it was decent money as a side job so I didn't bother asking. Possibly similar motives to this article?
> Very few of members of the foreigner community are fluent in Korean. Foreigner access to Korean BW discourse is a contradicting concept: if you speak Korean fluently, you have no reason to be in the foreigner community, as it only has access to material that is strictly inferior and more limited. For this reason, Korean-speaking members in the foreigner community are exceedingly rare.
I can vouch for this in general - after becoming fluent I've stopped looking up anything related to Korea in English because the quality of information is much worse. I'm sure the same holds for other languages and places.
I was a reasonably competitive BW player until the Korean teams arrived on the scene. I'll always appreciate how they elevated the level of gameplay. Really nice guys, too, I learned a lot from playing with them, and it was fun talking strategy via chat to the best of our mutual linguistic abilities. Good memories. I would have absolutely loved something like this project back in the day.
Warms my heart to see effort put into a beloved game. Just this week I watched a YT video of a sc custom game where the players were discussing whether its worth the effort to translate in-game korean-language content. Its an old game that is played by a niche community in north america. The majority of custom games are created in korean and never get to translated for the small number of north american players that would be interested.
> "Starting off the most affluently" is an awkward and verbose way to say "this is an economic opener/build."
Why would it be awkward and verbose?
Nit-picking, but a high signal to noise ratio is desirable, indicating low levels of noise compared to signal, not the reverse.
Impressive project, and I always love reading about the communities that form around competitive games.
It feels kind of sad to admit as a chess player, but "Najdork variation" is one of the funniest typos I've seen.
> Najdork variation
I think OP doesn't like the Najdorf Sicilian... or is this some meme opening I don't know about?
I wonder how well AI audio generation would work here, to produce a voiceover video like the original input.
Pretty cool and shows a clear issue I’ve seen across any LLM. the language and grammar is so formal/robotic
Neat. I wonder what Google Translate uses these days and if its gotten or will receive an update to a new LLM.
As the author points out, this does seem like exactly the kind of language problem that LLM‘s ought to excel at, and I love that moment of discovery when the testers were so busy discussing the content that they forgot to focus on the accuracy of the translation!
ok but where are the vods?
I was able to understand the Google Translate version well, but I am very familiar with the intricacies of BW and zerg 12hatch openers.
Chatgpt and Claude did an incredible job translating the korean text:
Claude: