Related link: http://www.newscientist.com/news/news.jsp?id=ns99991192
Computer scientists at OASIS have formed the HumanML technical committee, whose mission is to create an XML spec for cross-cultural meaning. HumanML is an extension of the “Semantic Web” concept applied to actual human language, where authors can code the subtext to their words, and have that subtext be preserved across translations of that text.
Think of it as a DTD for emoticons… ;)
From the article:
“With HumanML, web pages could be used to express meaning to someone who speaks an entirely different language. Humour is often lost during translation, but could be tagged as funny in HumanML. This could then be represented in a HumanML-enabled browser as pictures, sounds or words.”
“The creators of HumanML believe that eventually the language might make an ideal communications tool for virtual reality or artificial intelligence programs.”
Linguists have struggled with machine translations for decades, and yet Babelfish still only produces vaguely similar translations of web pages. Could HumanML be the answer?