Google’s new addition to its algorithm, Knowledge Graph, is set to add a ‘human touch’ to search results and provide instant answers to queries.
Hailed as “the next generation of search”, Knowledge Graph works by using an intelligent model to understand and translate search queries into “real-world entities and their relationships to one another”. As Amit Singhal put it on yesterday’s Google blog: “things, not strings”.
Take, for example, the query “Taj Majal”, which has the potential to mean a variety of different things to different users. Knowledge Graph will use what it knows about the keywords in a broader, deeper way, to provide a more relevant result. Singhal explains:
“You might think of one of the world’s most beautiful monuments, or a Grammy Award-winning musician, or possibly even a casino in Atlantic City, NJ. Or, depending on when you last ate, the nearest Indian restaurant.”
The feature will move away from simple keyword matching, and move towards matching places, people, ideas, history, art, landmarks, objects and whatever else correlates with the search phrase you have entered, in a more comprehensive way. It does this by utilising public information sources, like Wikipedia, Freebase, and the CIA World Factbook, and by drawing on the 3.5 billion facts it knows about the 500 million objects on its database. This means you’re more likely to get a targeted, local, relevant response to a potentially ambiguous query.
It will also read your mind.
Well, at least, anticipate your next question. For example, results for “Tom Cruise” will also include 37% of answers for the queries that then usually follow that particular search.
Knowledge Graph will initially only be available in the US, but will soon be rolled out across the world.
The ultimate goal, according to Singhal, is to be able to answer complex, specific (or even abstract) questions, such as: “What are the 10 deepest lakes in Africa?” and to develop a search engine that truly answers questions like a human being.
You can find out more about Knowledge Graph in this video: