(By Robert Hof)
“This update to the algorithm focuses more on ranking sites for better relevance by tapping further into the company’s Knowledge Graph, its encyclopedia of 570 million concepts and relationships among them. (For example, there’s a Knowledge Graph “card,” or information box, for the Eiffel Tower, and Knowledge Graph knows it’s a tower, that it has a height, that it’s in Paris, etc., so Google can anticipate you might want to know some of those facts.) Caffeine was more focused on better indexing and crawling of sites to speed results.“
Google has updated its core algorithm that controls the answers we get to queries on its search engine in a bid to make them work better for longer, more complex questions.
The update, code-named Hummingbird, is the biggest change to the underpinnings of the world’s leading search engine since early 2010, when Google upgraded its algorithm to one it called Caffeine. Google made the change about a month ago, it announced at a press event in the garage of the Menlo Park (Calif.) house where Google started. The event also celebrated the 15th anniversary of Google’s founding, which is today.
Most people won’t notice an overt difference to search results. But with more people making more complex queries, especially as they can increasingly speak their searches into their smartphones, there’s a need for new mathematical formulas to handle them.
This update to the algorithm focuses more on ranking sites for better relevance by tapping further into the company’s Knowledge Graph, its encyclopedia of 570 million concepts and relationships among them, according to Amit Singhal, Google’s senior VP of search. (For example, there’s a Knowledge Graph “card,” or information box, for the Eiffel Tower, and Knowledge Graph knows it’s a tower, that it has a height, that it’s in Paris, etc., so Google can anticipate you might want to know some of those facts.) Caffeine was more focused on better indexing and crawling of sites to speed results.
After the event, Scott Huffman, a key engineering director at Google currently working on natural language, told me that part of the impetus for the change was that as more people speak searches into phones, they’re doing so in a more natural way than they type in queries–which is to say more complicated. So Google’s search formulas needed to be able to respond to them.
Partly that is through even great use of the Knowledge Graph, so obvious discrete terms can be identified quickly. But it’s also interesting that although queries are getting more complex, that doesn’t always mean it’s harder to find the right answers. The more terms people use, Huffman says, the more context Google can divine. So those extra words, even if they’re in a more complex query, can give Google better information–but only if the algorithms are adjusted to be able to recognize the relationship among those terms.
Ultimately, he says, “we want to get to a natural conversation” between people and Google search on whatever devices they’re using.
You can read a few more details on Google’s blog post and see other accounts on Techmeme. Here’s my live account of the event:
Google is announcing some updates to its search engine at a press event in Mountain View this morning.
We just got bussed to 232 Santa Margarita, Menlo Park–the house where Google started. So it must be something special! Inside the house are Googley lights and a neon Google sign.
It looks like they’re going to celebrate the 15th anniversary of Google. I think I overheard that cofounder Sergey Brin might come. Update: No Sergey. Oh well.
And now we all pile into the legendary garage! Susan Wojcicki, Google’s senior VP of advertising and commerce, who owned the house when she rented it to founders Larry Page and Sergey Brin, tells us before the event that they were working all the time. There’s a toolbox, a drill, a Skilsaw–no doubt for show–along with a small stage with screen and podiums. And of course a colorful Google bike hanging on a wall.
Wojcicki starts by talking about the early days. It’s a little strange to be doing a press event and seeing TV cars in front of my own house, she says. She lived here seven years. Larry and Sergey were at Stanford in 1998 and were looking for office space, she says. She was worried about paying the mortgage, so the deal with her husband was that if they were going to buy it, they’d rent part of it out.
Most things in the house are original–including the blue carpet under our feet in the garage. I didn’t really know what to think about the company, she says, and by the time seven people were working and living here, it was getting time to find other space.
Before I knew it, says Wojcicki, who was working at Intel at the time, all my searches were on Google. I realized how important Google became to me, she says, and realized she wanted to be part of Google.
She still had to interview, though she didn’t manage to join while they were still in the house, which would have meant a very short commute. We all had Palm Pilots, she says, but a lot of things are the same today. Three things in particular:
The first one is a commitment to building great search experiences for users, she says. She recalls one example of someone who suspected a heart attack was coming on, searched Google for symptoms, and realized the person needed medical help.
Second, we wanted to be a global company, she says. If you go into the back room in the house, there was a whiteboard that said “Google’s worldwide headquarters.”
When I got to Google, I asked, who should we market to? They said, “Everybody.” She said she didn’t learn that in MBA school.
Third, there was a focus on thinking big. We’re still thinking about making technology that will change people’s lives, she says. So even though a lot of things are different at Google, a lot of things are the same.
Now comes Amit Singhal, SVP of search, to trace the history of search. We are just getting started, he says. Search has been foundational to the development of online, he notes. Search has been the main catalyst by making all this growth on the Web findable.
Google will keep reinventing itself to give you all you need for a simple and intuitive experience, he says. At some point, pulling out a smartphone to do a search will feel as archaic as a dial-up modem.
Some visible milestones along the way in search: spell-checking in search (2001), the concept of synonyms in a search (2003), autocompletion of queries (2005), universal search on all kinds of topics in one interface (2007), Google Instant to save several seconds per search (2010), Knowledge Graph to understand concepts and not just words (2012), and most recently voice search and Google Now, the predictive search service.
So what’s next? Singhal says Google should answer your questions, have a conversation in a natural way, and even anticipate what you might want to know.
First, on answers: Singhal touts the Knowledge Graph again. Shortly, you will see a couple of new ways to generate answers and get better knowledge. Second, you’ll see better voice search. And third, Google has been working on a new design language to anticipate what you want better.
OK, now to what Google is currently working on. Tamar Yehoshua, VP of search, is going to get more specific. She does a voice search, “Give me pictures of the Eiffel Tower,” and they appear. “How tall is it?” The answer is spoken. “Show me pictures of the construction.” “Here are some matching pictures,” the voice answers along with photos on the screen. The point here: The service knows what you’re searching on, so you don’t have to keep repeating “Eiffel Tower.”
Another example returns her request for Impressionist artists, with pictures. Then she clicks over to Picasso, getting one of the Knowledge Graph’s “cards” with basic info. There’s also a slide show at the top showing his paintings chronologically.
So the Knowledge Graph is great for exploring topics, she says. But Google also wants you to be able to do more complex queries. She had a recipe for coconut oil that she wasn’t sure was OK for someone’s food allergies. She asked for a comparison of coconut oil and olive oil–all kinds of links on that.
Google is coming out with an updated search app on Apple’s iOS. We’ve also simplified the home page for Google search, with voice search front and center. “Remind me to get olive oil when I go to Trader Joe’s,” she asks, and gets a map of the closest one. What about Safeway instead? Same thing. She sets a reminder so she’ll get pinged when she gets there.
A few of the specific tweaks Google is making in this vein are giving mobile search a more consistent look across smartphones and tablets, allowing you to compare two related things in a search (like coconut oil and olive oil), and releasing a new iOS 7 app, which lets you sync reminders on iPhones and Android phones.
Singhal returns with one more update on a change Google recently made. We have changed Google’s engines mid-flight again, he says. We probably didn’t notice, he adds. Basically, it updated the algorithm to something Google calls Hummingbird. It affects 90% of searches worldwide. He’s not specific about what it does, so it sounds like further improvements in what they just talked about.
Now for questions:
Q: How big a change is Hummingbird? Singhal says it’s as big as the change to the last algorithm, known as Caffeine. It happened about a month ago.
Q: How specifically is Hummingbird better: Singhal says it’s essentially to better answer the much more complex queries people are making. It impacts all kinds of queries, but far more effective on long, complex questions that we’re getting many more of now.
Q: Examples? Hard to be specific, but essentially, with more complex queries, the algorithm can better understand concepts vs. words as well as relationships between concepts.
Q: How is this different in nature from Caffeine? This is clearly more focused on ranking sites better for relevance, while Caffeine was more focused on better indexing and crawling of sites.
Hummingbird gave us an opportunity to rethink how we can reuse all these new services to improve search results.
“Opinion pieces of this sort published on RISE Networks are those of the original authors and do not in anyway represent the thoughts, beliefs and ideas of RISE Networks.”