Last month, Google CEO Eric Schmidt said during the company’s earnings call that Google had implemented about 120 search quality improvements during the third quarter as it moves toward its ultimate goal: “We want to get to the perfect search engine.”
No one asked him to elaborate on that lofty goal, so when IDG News Service recently had a chance to interview Marissa Mayer, Google’s vice president of Search Products & User Experience, we promptly asked her to explain what Schmidt meant. She also talked about what keeps Google ahead on search, how the company views semantic technology and what’s next in its Universal Search efforts to combine links to a variety of file types — news articles, images, videos, books, maps — in a single results list.
An edited transcript of the conversation follows:
IDG News Service: What is the perfect search engine? If you had a magic wand and could create it, what would it look like? What would it do?
Marissa Mayer: It would be a machine that could answer that question, really. It would be one that could understand speech, questions, phrases, what entities you’re talking about, concepts. It would be able to search all of the world’s information, [find] different ideas and concepts, and bring them back to you in a presentation that was really informative and coherent.
There are a lot of different aspects of research that need to go into building that search engine. You need to understand speech. You need to understand images. You need translation, so you can find the answer regardless of what language it’s written in. You need a lot of artificial intelligence to be able to analyze what information is relevant and synthesize it. You need a great user interface and user experience to put it in context. And you probably need a certain amount of personalization, so the search engine relates to the person, to their background, what they already know about, what they looked for last week.
IDGNS: At the user interface level, Google gets criticized by its competitors constantly for what they pejoratively disdain as Google’s “10 blue links” results page. They say Google is old school, that its paradigm of search is inefficient and inconvenient. How do you respond to that kind of criticism?
Mayer: I’d point to the fact that Universal Search was really a watershed moment in this. You get diagrams, pictures, blogs, local information, books, news, all stitched into your search engine. While many of our competitors are still busy building small, vertical search engines where you have to remember they have them, we’re busy doing a very difficult computer science problem: How do you stitch all of these disparate mediums together into one coherent set of answers, and how do you synthesize all of that? We’re doing all of that because it’s better for users: Here’s the tool and it gives me what I want, regardless of what format it came in.
We have two, three, five changes every week that are visible to the end-user in the user interface. We don’t [publicize] the ranking changes. We are making changes to our ranking algorithm at the rate of two per day. Interestingly, some of our competitors haven’t made any changes to their ranking function for quite some time. Search needs to evolve: the user interface, the ranking function. It’s a process of making lots of small changes all the time and to constantly make things better.
IDGNS: What’s the status of semantic search at Google? You have said in the past that through “brute force” — analyzing massive amounts of queries and Web content — Google’s engine can deliver results that make it seem as if it understood things semantically, when it really functions using other algorithmic approaches. Is that still the preferred approach?
Mayer: We believe in building intelligent systems that learn off of data in an automated way, [and then] tuning and refining them. When people talk about semantic search and the semantic Web, they usually mean something that is very manual, with maps of various associations between words and things like that. We think you can get to a much better level of understanding through pattern-matching data, building large-scale systems. That’s how the brain works. That’s why you have all these fuzzy connections, because the brain is constantly processing lots and lots of data all the time.
IDGNS: A couple of years ago or so, some experts were predicting that semantic technology would revolutionize search and blindside Google, but that hasn’t happened. It seems that semantic search efforts have hit a wall, especially because semantic engines are hard to scale.
Mayer: The problem is that language changes. Web pages change. How people express themselves changes. And all those things matter in terms of how well semantic search applies. That’s why it’s better to have an approach that’s based on machine learning and that changes, iterates and responds to the data. That’s a more robust approach. That’s not to say that semantic search has no part in search. It’s just that for us, we really prefer to focus on things that can scale. If we could come up with a semantic search solution that could scale, we would be very excited about that. For now, what we’re seeing is that a lot of our methods approximate the intelligence of semantic search but do it through other means.
IDGNS: Universal Search was announced in May 2007. Is it considered finished now? Is it something that will always be a work in progress?
Mayer: It’s still a very living, breathing thing. Now we have multiple teams: We have a local [search] universal team, an image [search] universal team, the product [search] universal team. They’re all looking at how can we do an even better job ranking and triggering this content. When we launched it, it was showing in about one in 25 queries. Today, it shows in about 25 percent of queries. And we think there are probably times when those auxiliary [file] formats could actually help, and we aren’t triggering them on our results page. That’s something we need to continue to strive to do.