Rikai Does Chinese
December 17th, 2005 by MarkA few days ago, I went back to 台北 to hang out with one of my old co-workers one more time before he went back to America. It was pretty cool to hang out with him, and it turns out he’d been reading my blog. He said he liked the News in Chinese link I have up here, though it loads a bit slowly. The pop-ups with pinyin and English for every character make it possible for you to read Chinese news articles really quickly, even if you only know half or a third of the characters it would take to be literate. Then he mentioned that it would be nice if there were some java script tool that would do the same thing for you on any Chinese page.
Back when I was learning Japanese, there was a tool just like that: rikai.com. Rikai, or りかい (理解) means “understand” in Japanese. At that page you can either type Japanese into a box, or type in a URL of a Japanese page. Then, rikai.com generates a page that has the same text, but also javascript pop-ups with English and romanji for every word. It was an awesome tool. I used to use it to read several articles a day from major Japanese news sites like Yomiuri or Mainichi.
As I was talking with my friend, I realized something. Yomiuri translates Chinese too! I never cared about that before, while I was learning Japanese, and by the time I started learning Chinese I’d forgotten all about Rikai. I tried it out on an article from the 工人日報, and I found it to be even more convenient than newsinchinese.com since it loads so much faster. Since it is also usable on nearly any Chinese page, there’s no contest between the two tools. I’m removing my newsinchinese.com link, and replacing it with rikai.com. To use rikai.com for Chinese, go to the main page and select “Chinese to English” from the drop down menu on the right. Then, you can either start clicking on news articles on the bottom or enter a URL into the box to the upper left.
To see the article I was reading in the picture above, follow this link.
:
December 19th, 2005 at 9:27 pm
A friend passed word of your post, and since I’m involved with Adsotrans I thought I’d leave a reply. I personally think Rikai is a great service, so just want to comment briefly about some important differences between it and us:
First of all, there is no question Adso is slower than Rikai. This is because Adso tries to be more of a translation engine than an annotation one. It unifies dates, names, numbers, foreign loan words, etc., and tries to disambiguate texts to offer the most likely pinyin/gloss for any entry rather than all possible definitions. Users may not notice this until Adso screws up and confuses Pakistan (巴) with Palestine (巴), but it means that the software requires backend processing time analogous to other translation engines on the web.
It also means pushing beyond flat-file wordlists like CEDICT and trying to develop a richer database with extensive word-level information — a database that knows Beijing and Toronto are both cities and that 布莱德 and 彼特 are both personal names. We put a lot of effort into actively expanding the size of our backend database and try to encourage users to help out. It does make a difference in quality. In areas where users have added terminology, for instance, it isn’t terribly uncommon for Adso to best systems like Google at machine translation.
I’m not sure what problems there are with speed issues with NewsinChinese, although the entire project is based off a single server running in Beijing (all I can afford) and we’ve been adding services for some time. Feedback is always welcome at the site in any event, or at our sub-forums in the computing section of Roddy Flagg’s http;//www.chinese-forums.com
Cheers,
–david
December 19th, 2005 at 11:49 pm
This is interesting. I really appreciate the reply. I hadn’t realized that newsinchinese.com was run by Adsotrans. Your site says that Adso uses “artificial intelligence”. Are we talking neural nets, GE, or something else?
As for the speed problems, I’ll do my best to explain. Each page load is excruciatingly slow. Maybe this is since your server is in Beijing, and mainland China as a whole has very very insufficient bandwidth between itself and the rest of the world. It’s possible that for users currently in mainland China, the service is quite fast. For that matter, one reader in Shanghai told me that http://toshuo.com/ is blocked in China. Have you been able to access my page without a proxy? In any case, I and all of my other friends learning Chinese in Taiwan have found page loads for newsinchinese.com to be ludicrously slow.
This problem is further compouded by the fact that after loading up the page and clicking on an article, one must click “more” and wait once more for the page to load. Then, it’s necessary to click on “full” in order to get the whole article. After the surfer girl appears, the reader must then click on “ready”. I just went through this process and it took 18 seconds to load the main page, 5 seconds to load “more”, 11 more seconds to load “full”, and finally 2 more seconds after clicking on “ready”. The service loses a lot of its value as an extensive reading tool if it takes 36 seconds of loading per page of news.
December 20th, 2005 at 12:11 am
Travelan, can adsotrans translate any web site? I mean, is there a way to put in a URL of a Chinese page and get it annotated?
December 26th, 2005 at 1:47 pm
Sorry for the delayed reply. Tushuo is blocked in Beijing as well and I wasn’t even sure my comment came through last time.
On the technical front, the engine is a modular and extensible rules-based system. It sounds like the speed issues with newsinchinese at least result from our having a Beijing server, so there isn’t much we can do about that save than say that we’d be happy to get a mirror set-up if anyone is interested in hosting something abroad.
And Jim — it isn’t advertised, but if you stick a URL that beging “http://…” into the annotation box on the advanced page the server will process it. Just be sure to specify the encoding of the remote page correctly or you’re likely to get gibberish.