Quite out of the blue and without notification of it’s launch as far as I’ve been able to find, Google seem to be exposing semantic data in their global search results.
Try searching for ‘What is the capital city of England?’ or ‘Who is Bill Clinton’s wife?’ and you’ll see sourced direct answers returned at the top of your search results.
It’s hard to tell if these direct results are actually semantic expressions or just presented to appear that way – in the expected Semantic triple of subject-predicate-object. The list of sources definitely don’t structure their information with semantic expression, so perhaps quite an amount of logic and natural language processing is being done on Google’s part to process non- or semi-structured data.
I’ve tried to find out before what Google have been up to concerning semantic technology but found little. The coverage over at ReadWriteWeb reports that neither they or their Semantic Web contacts had heard or seen anything about this before, but the community feedback suggests there’s been variations of this for some time – including a three year old Google program called ‘Direct Answers’ – but none of the coverage of that program offers the kind of examples we’re seeing here.
Marshall Kirkpatrick points to a blog post of Matt Cutts, Google search algorithm engineer, but it seems to be a dead link now. Though trailing through Google’s caches, it seems to find him quote:
Many of the data points are being pulled in from the structured part of Wikipedia entries, which is interesting. Other sources are wide ranging, from a license plate website to Jason Calacanis’s Mahalo.
If Google are constructing semantic data from semi-structured or non-structured source data, then there’s undoubtedly some quite powerful semantic processing technology in place. I highly doubt this will be the final product of their development with such technologies, simply the first we’ve noticed – most likely why it’s slipped under most people’s notice.
The inaccuracy is also an issue. Try searching ‘Who is Bob Dylan’s wife?’ – and you’ll see Sara Lownds (his ex-wife) returned. Seeing these direct answers reminds me of True Knowledge.
Even their example questions though, are far more complex – for example, ‘Who owns Universal Studios?‘, ‘Is the Chrysler building taller than the Eiffel Tower?‘, ‘What is the half life of plutonium 239?‘.
More importantly, if it doesn’t know that answer, it won’t ‘guess’ – it’ll tell you it doesn’t know and ask you to deconstruct your query in order to expand it’s knowledge base so it can find the answer later.
As Marshall says, this is all speculation based on limited observation – and low visibility of Google’s development. Hopefully there’ll be more soon!