Google rolls out Knowledge Graph to all searches using English
Intelligently determines what the user wants based on his search history
THERE’S that memorable scene in 1986’s Star Trek IV: The Voyage Home where the crew of the starship Enterprise has travelled back in time from the 23rd to the 20th century. Chief Engineer Scotty (James Doohan) attempts to use a Macintosh.
Facing the screen, he speaks to it (of course): “Computer? … Computer!”
His crewmate Dr ‘Bones’ McCoy (DeForest Kelley) looks as puzzled as him, then notices the mouse, which he hands over to Scotty. Scotty goes, “Ah!”, then speaks into the mouse, “Hello, computer.”
It doesn’t work. The 20th century engineer they’re trying to impress says, “Just use the keyboard” to which Scotty responds, “Keyboard. How quaint.” (Click here to refresh your memory).
Having conversations with your computer – beyond cussing it out when it crashes – may no longer be a Hollywood pipe dream if Google’s senior vice president in charge of search, Amit Singhal (pic), has his way.
“We’ve taken baby steps towards a childhood dream of mine – to build a Star Trek computer,” he told Asia Pacific journalists during a video Hangout call to announce the latest feature on Google search – the Knowledge Graph, which provides more intelligence and context to your searches.
The Knowledge Graph was rolled out to US users in May, and from today onwards, will be available to all users doing Google searches in the English language. When searching for people, places or things – it works on entities and not abstract queries – a summary appears on the right hand with the most relevant searches and summaries. For example, if you’re doing a search on P. Ramlee, your little Knowledge Graph window will contain background information and his birth date, whom he was married to, his albums and details on his famous movies.
More importantly, it lays out the possible objects of your search if there is more than one possibility. Amit’s example was a search for Chiefs – which in the United States may refer to a football team, but which is a rugby team in New Zealand.
The company says that Knowledge Graph understands real-world entities and their relationships to one another, which means that Google (the search engine) can understand your search query better and give you more information you’re likely to need, right on the results page.
“We are working on making it available in other languages, but it’s a real challenge,” admitted Amit, who declined to give specifics or any timeframe. He noted that within the English language itself, there are regional and cultural differences in what may be the most relevant search results, citing the above ‘Chiefs’ example.
Another example he quoted more than once during the Hangout was “Andromeda,” which could refer to the galaxy, the TV series or a Swedish band! (Click to enlarge example).
“Over the past two years, we've created a database of structured knowledge containing 500 million different people, places and things, with 3.5 billion defining attributes and connections between them,” he added.
How exactly Google is doing this is a well-kept secret, but Amit said that part of it is determining what the user wants based on his search history as well.
In its on-going evolution towards Universal Search, Google also announced a trial which would merge search with Gmail, so if you keyed in “Flights,” the Knowledge Graph would also parse your Gmail to find your airline's email notification containing your flight information, for example.
To participate in the trial – the company is opening up to only the first one million Gmail users who request to take part – go here. This trial is also only accessible on https://www.google.com/ in English and for @gmail.com addresses (not Google Apps accounts). “Due to limited capacity and initial setup, you will receive an email confirmation from Google when your account is ready, but note that we may not be able to activate the experiment for everyone,” the company adds.
Tim Berners-Lee, the inventor of the World Wide Web, coined the term the “Semantic Web” to describe what he sees as the next evolution of the Web, into "a web of data that can be processed directly and indirectly by machines."
But despite on-going attempts by the W3C (World Wide Web Consortium), Berners-Lee has been quoted as saying that "this simple idea ... remains largely unrealized."
“We’re making the vision of the Semantic Web a reality, on a very large scale,” said Amit however.
How far are we from a Star Trek-reality where computers can converse intelligently with you, understand what you’re really asking it, and give you intelligent responses and recommendations – indeed, advice? (And make it Kirk-proof so that smarmy young captains can’t cause it to self-destruct by posing a logical conundrum or two.)
“This is just the first step – there are several others that need to be taken,” he admitted. “Speech recognition and natural language understanding are real challenges, but I intend to achieve my childhood dream,” said Amit.
While you’re at it, ponder this: Speech-to-text software => Parse through Google Translate => Text-to-speech software = Universal Translator? Go to it, Google!