Google announced a new feature of its search engine yesterday that presents information, not just web links, in response to search queries.
The feature, called the Knowledge Graph, works by analysing historical search queries to identify useful information a visitor may be looking for. For example, a search for an author will return a list of books they wrote, while a search for an architect will return buildings they design.
It also displays information from public websites Google has indexed, such as Wikipedia, Freebase and the CIA World Factbook.
The Knowledge Graph itself is an ontology that represents more than 500 million objects, including sports teams, buildings and works of art. These objects are linked together by over 3.5 billion facts and relationships.
As engineering SVP Amit Singhal explains in a blog post, identifying these objects allows Google to prompt the user to clarify their search. For example, if a user searches for ‘the Taj Mahal’, the search engine can refer to the ontology to discover that this could mean "one of the world’s most beautiful monuments, or a Grammy Award-winning musician, or possibly even a casino in Atlantic City, NJ."
Google’s search results page will let users indicate which meaning of their search term they are interested in by clicking on links that filter out other meanings.
Many of the features of the Knowledge Graph, which Google has begun to roll out in the US, are already in use by ‘computational knowledge enginer’ Wolfram Alpha.
Both engines, for instance, output the height of the Taj Mahal when that string is searched for. However, the answers differ: Google says the palace is 561 feet high but Wolfram Alpha says the height is 243 feet.
The difference highlights the difficulty of providing accurate, automated semantic search results. Google’s answer appears to be taken from the Taj Mahal’s Wikipedia page, and actually refers to the building’s height above sea-level.