Computers vs. Humans (Is Hal on our heels?)

by desa on March 29, 2011

Luckily for human translators the continuing problems with accurate machine-generated document translation mean that the technology continues to be considered “infantile” or “burgeoning” all of which means our jobs are safe for now. That’s not to say that there aren’t some fascinating developments.

Rohini Shrihari, a Computer Scientist at the University of Buffalo says translations of Urdu, the language of Pakistan and Muslims in India, are too literal, especially in the context of social networking. So she developed a natural language program that, she says, can understand the context of conversations and identify the people, places and things being talked about on Twitter for example. The computer will produce a document translation of the conversation that the user can scroll over with a mouse, as you would any document. If there were a negative connotation in the document that part of it would glow red; in contrast it would glow green if something were being discussed in a positive way.  Shrihari’s research is of interest to the Pentagon, which funds her company, and is interested in any information her Twitter scavenging might yield. “In Twitter posts and tweets and so on, if there’s specific factual information that’s being mentioned — they want that extracted,” Srihari says. “There’s also definitely an interest in sentiment and opinion mining.”

Ortsbo, a Toronto-based company, promises instant, real-time translation in 50+ languages for Facebook, Twitter and other social media. However, for business purposes one might want to choose one’s (translated) words more carefully. And of course there is Google Translate. It’s latest apps for iPhone and Android smartphones promise new (although limited) speech-interpreting powers. Again, business customers should probably use these at their own risk.

However, there is one important area that is still severely hindering the rise of the machines: Slang! Computer Scientist Shrihari reports her program is having some problems translating Urdish, a mashup text language that, as the name suggests, is part-Urdish, part-English. So Jibberish translators are probably safe for now.

Next post: