Answering questions Google can't

GOOGLE magic doesn't always work.

"I always use Google personally and I really like it, but there are some things that I can't find," said Menno van Zaanen, whose field is computer linguistics.

Not long ago van Zaanen and friends were wondering about the origins of the apparently redundant practice of announcing "The End" of a book just as it runs out of pages.

"If you search for 'The End' in Google you won't find the answer," he said.

At Macquarie University Dr van Zaanen and his team are building a different kind of search engine, one that seeks specific answers to plain-language questions.

"[In AnswerFinder] you type in the entire question ... an example would be: 'How many Australians live outside Australia?' Instead of giving a whole list of documents, which Google does, the system not only analyses the question but tries to find the exact answer. In Google, you can type in a question but it will just throw away some words because they're too common, but these words may actually be important for the answer."

AnswerFinder is an exercise in computational linguistics and deploys a tool known as a parser, just as generations of schoolchildren once used to parse sentences, breaking them down into parts of speech. Of course, computers sometimes struggle with the subtleties of language.

"Analysing language in general is really, really hard [for computers]," Dr van Zaanen said. "Somehow we're quite good at it but we just don't really understand why we are so good at it."

Others making up the research team at Macquarie's Centre for Language Technology are project director Diego Molla and PhD students Luiz Augusto Pizzato and Elena Akhamatova.

AnswerFinder was still in development but Dr van Zaanen said it was the kind of thing that could work for companies in information technology, telecommunications or media, companies whose users needed to search larger databases of documents online.

As a corporate website amenity AnswerFinder is no putative rival to Google with its web-wide reach and vast processing power.

Even so, corporate websites with search facilities these days typically do use Google or other systems that require users to have some skill (or luck) in devising keywords.