Thoughts on Bing and Search Engines of the Future, From UW Computer Scientist Dan Weld
Dan Weld spends a lot of time thinking about the Web and how to get the best information out of it. Weld is a professor of computer science and engineering at the University of Washington and a serial entrepreneur, having co-founded Netbot, AdRelevance, and Nimble Technology. He is also a venture partner with Seattle-based Madrona Venture Group. Some of his research projects include customizable software interfaces and ways to improve information finding on Wikipedia. He is an expert in Web search, information extraction, and adaptive user interfaces, so he seemed like the perfect person to ask: what is the big deal with Microsoft’s Bing, anyway?
In a recent interview, Weld talked about the improvements Bing has made over Google and other current search engines (only slight), the future of Web search, and a hint at a project he is working on that he thinks could change the way we find information online.
The following is an edited version of our conversation.
Xconomy: Let’s start with your general impressions of Bing.
Dan Weld: I think it’s a nice, if small, advance. Some of the things they’ve done in this relaunch are primarily architectural and will support their plans for the future. In terms of what is actually available right now, the biggest change is integrating vertical search in a uniform way.
X: What does that mean?
DW: When you think of search, everyone thinks about Google, and maybe Yahoo, but there are many other kinds of search, like people searches (you can think about Facebook as a people search), travel search (Kayak and Expedia), health information sites. You can think about Wikipedia as providing a search for encyclopedic information. Shopping searches—Amazon is great in part because it makes it easy to find so much information about the products, and reviews of products. All of these are examples of vertical search experiences. Instead of having wide coverage, you have a better experience within a narrow range.
Bing has tried to marry those things into an integrated wide search experience. All of the engines have been doing this. If you do a search on Google for a movie, you might see information about show times and trailers at the top, for example. Bing has gone further in some directions than people have gone before, in this aspect. If you look at their tabbed pane, it lets you look at different kinds of information right there.
X: But doesn’t Google do that too?
DW: All the engines are trying to do it, but the way Microsoft has done it with Bing is somewhat better than what Google has done. With the shopping tab, you get a faceted interface, meaning you can narrow your search using categorical information, restricting yourself to a particular brand or price range. Those facets are specific to the object you’re searching.
X: So if this is just a small change, what kinds of big changes can we expect to see in the future?
DW: Lots of people are happy with search today, but that’s because they set their sights too low. I think search is going to change enormously in the future, and I think it’s going to do so by adding increased natural language understanding, machine learning, and artificial intelligence into the interface.
So what’s it going to look like? Today, you issue a query and you get an ordered list of pages that you might want to read, because they might answer your question. I think what we’ll see more of in the future is the ability to ask a question and have the search engine actually answer your question. The simplest way to describe this only works for old geezers like me, and that is the example of the ship computer on the original Star Trek series, where you could ask it a question and it would tell you the answer. I think eventually we’ll see technology like that, but it’s going to take the engineers at Microsoft, Google, and lots of universities to get there.
Another way I see search going, and this may be closer than the Star Trek computer, is what I would call entity-oriented search. Usually when somebody issues a search query, they have some object in mind, and if the system can interpret what that object is, it could go further in answering your question. Search engines do that most effectively today when it comes to shopping. When you search for a product, you want to know its features and the best or cheapest place to buy it. But on factual queries, it could be similar. So if someone enters “Seattle,” maybe they’re interested in historical information about Seattle, or maybe they want hotels and restaurants, and having the information displayed to the user in terms of those aspects is going to be much more useful than giving a list of pages.
One final direction is increased personalization. You don’t really see that in the Bing interface right now, but I know one of the major changes they’ve done is to make it capable to add increased personalization. Google and the others are aiming at personalization too. A large percentage of queries that any one person does are re-queries, meaning they’re either the same search terms they’ve used in the past, or different search terms, but trying to get the same piece of information. So if the engine keeps track of what I’ve searched before, it can do better. And the challenge is how to do that quickly.
X: So is Bing getting so much attention because it’s Microsoft, or because it’s really revolutionary?
DW: There’s a third option, of course, and it’s that they’re spending something like $80 million to promote the thing. I think it is better in a lot of ways, but I wouldn’t call it revolutionary. Soon I hope I can tell you about some of the research we’re doing which I think will eventually revolutionize search. Our “Intelligence in Wikipedia” project has developed new methods which will analyze the Web to build a huge knowledge database. Initially, the interface will look like a relational database [a database that draws connections between related groups of data], but before long we will provide a complete question answering system.
Trending on Xconomy[toplytics period="week" numberposts="3"]
By posting a comment, you agree to our terms and conditions.