Microsoft Corp. researchers in Beijing are using data mined from the Web to enhance an online Chinese-English dictionary and language-practice service, a technique that could one day be used in similar tools for anyone learning any language.

Engkoo, at www.engkoo.com, is written in Chinese with two characters meaning 'English' and 'vault.' It has a core of professionally produced translation data that Microsoft draws from sources such as existing dictionaries, which are used under licenses from their publishers. In Engkoo's database, that content is mixed with data that Microsoft finds through other means, including sweeping the Web for sites with parallel Chinese and English versions. Microsoft machines align the two websites -- followed by their paragraphs, sentences and individual words -- then assign a quality ranking to the resulting translation and file it away. Engkoo is a finalist in this year's Asian Innovation Awards.

When an Engkoo user types a word or sentence into the website's input bar, in either Chinese or English, the site draws on statistics from its vault of data to translate it. It also displays samplesentences that use similar words and, in many cases, links to where it found them.

Sourcing translations from the Internet can help the database keep up-to-date with evolving language, such as new colloquial or technical terms, the researchers say. Engkoo users can also report translations that look wrong. Human editors fix any serious errors and improve the technology where possible to prevent the problem from recurring. 'This is a system that gets smarter over time,' said Matt Scott, a development group head at Microsoft Research Asia. 'We want translation to reflect the Web.'

Statistical machine learning for translation services is now widely researched and is also used by some other websites, like Google Translate. But the researchers behind Engkoo are also tapping other technologies to expand their website's range of language-practice tools. For many English samplesentences on the site, users can listen to audio dictations, which a machine generates based on collected audio files of native English speakers talking. The dictations are meant to imitate human inflections, though their up-and-down swings don't yet match those of a natural voice.

Microsoft's researchers are also working on a video dictation feature for Engkoo. The few videos already on the site were created in a similar way to the audio dictations -- by a machine drawing on sample video of an English speaker talking. The goal is for users to be able to watch and learn from the lip movements of a native speaker dictating any sentence, even though each video is machine-generated.

Since tongue movements are also crucial for pronunciation but normallyhidden from view, the researchers are gathering ultrasound data that could generate a parallel set of videos on Engkoo. One option is to turn the black-and-white ultrasound footage into more appealing cartoon animation showing users how a native speaker's tongue moves while speaking a sentence, said Frank Soong, a Microsoft Research Asia principalresearcher.

Engkoo was launched last year and gets more than four million visitors a month, according to Microsoft. Microsoft researchers are also developing a mobile Engkoo application for phones running a version of Windows, and apps for other mobile operating systems are also under consideration, Mr. Scott said.

The China version of Microsoft's Bing search engine already links to Engkoo and the researchers are talking with colleagues at Microsoft about other products where Engkoo could be integrated, said Eric Chang, director of technology strategy at Microsoft Research Asia.

Engkoo is free online. But a Microsoft Research spokeswoman declined to comment on whether Engkoo mobile apps will be free or could contain ads. People who use Engkoo may also use Bing services, which could eventually help drive advertisingrevenue for Microsoft.

Microsoft's researchers plan versions of Engkoo for other languages too, including Japanese and English. A version for English speakers learning Chinese is also a goal, but the company's research so far is focused on Chinese-to-English, Mr. Chang said.

Mr. Chang sees the mix of technologies being used by Engkoo as a step toward breaking down language barriers. Further technological advances may one day mean that, for instance, an English speaker in China could attend a university lecture in Mandarin and have no problem understanding the content.

'You could be sitting there, and then your mobile phone's actually doing real-time translation,' Mr. Chang said. 'Technology can really help in terms of reducing barriers in the use of language. So Engkoo is definitely a way for us to get more feedback on how people can utilize technology to do that.'

Owen Fletcher

• learning [´lə:niŋ] n.学习；学问；知识 (初中英语单词)
• parallel [´pærəlel] a.平行的 n.平行线 (初中英语单词)
• assign [ə´sain] vt.分配；指派；指定 (初中英语单词)
• sentence [´sentəns] n.判决 vt.宣判；处刑 (初中英语单词)
• sample [´sæmpl, ´sɑ:mpəl] n.样品；试样 vt.尝试 (初中英语单词)
• system [´sistəm] n.系统，体系，制度 (初中英语单词)
• research [ri´sə:tʃ] n.&vi.调查；探究；研究 (初中英语单词)
• reflect [ri´flekt] v.反射；反响；表达 (初中英语单词)
• working [´wə:kiŋ] a.工人的；劳动的 (初中英语单词)
• speaker [´spi:kə] n.演讲人；代言人 (初中英语单词)
• pronunciation [prə,nʌnsi´eiʃən] n.发音(法) (初中英语单词)
• hidden [´hid(ə)n] hide 的过去分词 (初中英语单词)
• principal [´prinsəpəl] a.主要的 n.负责人 (初中英语单词)
• application [,æpli´keiʃən] n.申请；申请书；应用 (初中英语单词)
• running [´rʌniŋ] a.奔跑的；流动的 (初中英语单词)
• consideration [kən,sidə´reiʃən] n.考虑；原因；体谅 (初中英语单词)
• director [di´rektə] n.指导者；....长；导演 (初中英语单词)
• comment [´kɔment] n.&v.评论；评注；注意 (初中英语单词)
• contain [kən´tein] v.包含；容纳；抑制 (初中英语单词)
• advertising [´ædvətaiziŋ] n.广告a.广告的 (初中英语单词)
• instance [´instəns] n.例子，实例，例证 (初中英语单词)
• actually [´æktʃuəli] ad.事实上；实际上 (初中英语单词)
• definitely [´definitli] ad.明确地；绝对 (初中英语单词)
• translation [træns´leiʃən, trænz-] n.翻译；译文；译本 (高中英语单词)
• sweeping [´swi:piŋ] a.掠过的 n.扫除；清除 (高中英语单词)
• translate [trænz´leit, træns-] v.翻译；解释；说明 (高中英语单词)
• technical [´teknikəl] a.技术的；专门性的 (高中英语单词)
• expand [ik´spænd] vt.张开；膨胀；扩大 (高中英语单词)
• imitate [´imiteit] vt.模仿；伪造 (高中英语单词)
• revenue [´revinju:] n.税收；收入 (高中英语单词)
• statistics [stə´tistiks] n.统计学；统计 (英语四级单词)
• drawing [´drɔ:iŋ] n.画图；制图；图样 (英语四级单词)
• gathering [´gæðəriŋ] n.集会，聚集 (英语四级单词)
• generate [´dʒenəreit] vt.创造；发生；引起 (英语四级单词)
• version [´və:ʃən, ´və:rʒən] n.翻译；说明；译本 (英语四级单词)
• eventually [i´ventʃuəli] ad.最后，终于 (英语四级单词)
• utilize [´ju:tilaiz] vt.利用；使有用 (英语四级单词)
• enhance [in´hɑ:ns] vt.提高，增加；增进 (英语六级单词)
• technique [tek´ni:k] n.技术；技巧；方法 (英语六级单词)
• innovation [,inə´veiʃən] n.改革；革新；创新 (英语六级单词)
• normally [´nɔ:məli] ad.正常情况下；通常 (英语六级单词)
• cartoon [kɑ:´tu:n] n.漫画；(电影)卡通片 (英语六级单词)
• speaking [´spi:kiŋ] n.说话 a.发言的 (英语六级单词)
• strategy [´strætidʒi] n.兵法；战略 (英语六级单词)