Tuesday, June 27, 2006

Valdes-Perez on the future of search

Raul Valdes-Perez (CEO of Vivisimo) is interviewed by Mark Roth of the Pittsburgh Post-Gazette. Some selected excerpts:
Where will search go in the future?

Many of the major companies are now investing in "personalized search," [Dr. Valdes-Perez] said, using a record of which sites a user has visited to help tailor their future searches .. [and offer] direct answers to people's questions ...

A big challenge there, he said, is to design a system that will understand what kind of information a person wants about a particular subject.

In general, an ongoing goal of search is to make it "smarter" so we can be "dumber" -- to develop software that will do the hard thinking for us.

"There's a quote I used to have on my home page at CMU that says 'Society advances by reducing the number of tasks that require thought.'

"You can accomplish much more by applying your thinking time to higher level stuff."
When I was at Amazon working on personalization, we used to joke that the ideal online bookstore would be very simple. It would display a giant picture of just one book, the next book you want to buy.

Yes, that is a bit too much. But, I think the point Raul makes is clear. People do not want to spend time thinking about searching. They want to get what they need and go.

The prize in search will go to those that help people get what they need quickly, effectively, and effortlessly.

2 comments:

jeremy said...

Valdez-Perez writes: In general, an ongoing goal of search is to make it "smarter" so we can be "dumber" -- to develop software that will do the hard thinking for us.

..and Greg responds: I think the point Raul makes is clear. People do not want to spend time thinking about searching. They want to get what they need and go. The prize in search will go to those that help people get what they need quickly, effectively, and effortlessly.

The point the Raul is making may be is clear, but the manner in which you are interpreting that point is not necessarily the same manner Raul is making that point.

Making search "smarter" is not the same thing as giving the user the "one right answer", i.e. allowing them to "get what they need and go" as you say.

I think the problem, Greg, is that you have a viewpoint of search as exclusively a "navigational" activity. You think of it as finding -the- web page or -the- book or -the- whatever. You think of all search as the TREC HARD task: The goal of HARD is to achieve high accuracy retrieval from documents by leveraging additional information about the searcher and/or the search context captured using very targeted interaction with the searcher.

However, search is not just navigation. There are all different types of searches. One type is "exploratory" search: "Search engines, bibliographic databases and digital libraries provide adequate support for users whose information needs are well-defined. However, when information needs are vague or evolving, searchers may benefit from interfaces that provide additional support, for example by enabling grouping of results and/or guided discovery processes."

When a user has a navigational information need, then yes, the way to make search "smarter" so we can be "dumber" is to put the most relevant result at the top of the list.

However, when a user has an exploratory information need, the way to make search "smarter" so we can be "dumber" is to improve the quality of the clusters, so that each cluster is more self-on-topic. Or to improve the quality of the query expansion "tools", so that the list of suggested terms is both more focused as well as more comprehensive.

Sometimes, due to the nature of the information need itself, it is actually "wrong" to deliver one single page as the "answer" to the search.. because one page is not what the user is looking for.

An example: suppose a person is interesting in learning more about the "causes of war". So the user goes to Google and types in that query. The top page contains a nice overview, but it is fairly general does not specify whether the user is looking for economic, political, social, racial, religious, or other causes of war. If the user wants to find out more about the most relevant result for each of these various causes, the user has to go back to Google, and be "smart" himself, by typing in 5 more queries, manually:

"economic causes of war"
"political causes of war"
"social causes of war"
"racial causes of war"
"religious causes of war"

So in addition to having to do all this himself, the user has to be doubly smart by having to self-enumerate every single one of these possible causes. The user might actually forget a facet, such as "industrial causes of war". In this case, the burden is on the user to be smart and figure out what to ask, rather than on the system to be smart and suggest it.

So in this respect, Google is actually quite "dumb" and forces the user to do all the work himself. Google forces the user to think up all the possible causes of war, and query for each one, one at a time. Even if the top 3-4 results do contain an enumeration of all the possible causes of war, Google still forces the user to read all these top pages, distill this knowledge themselves, and then come back and query Google one at a time.

A smarter search engine, on the other hand, could be (for example) one in which the user types in "causes of war", and the system comes back to him not just with web pages, but a list of possible query expansion terms.. such as "economic, political, etc."

In this second case, the search engine is doing all the hard thinking for us, by figuring out what the best terms are for expanding (more clearly defining) one's query. By smartly coming up with this list, the search engine lets the user be "dumb", and not have to think up or enumerate all the possibilities himself.

So I agree with you. People do not want to spend their time thinking about searching. But if they have a complex or exploratory information need (rather than a "what is the circuit city homepage?" information need) then today's most popular search engines do not help them at all. In terms of helping these latter searchers get what they need effortlessly, a philosophical approach like Google's actually requires more effort from the user than an approach like, say, Ask's or Vivisimo's.

I know, this is the same theme I keep beating my drum about. But, with no offense to you, it just makes me cringe every time I hear the attitude like the one you express with the perfect bookstore showing you the -one- next book that you want to buy.

I know that you're joking, but the philosophy behind what you say is still serious.. and very Googalian. And as such it completely ignores any kind of search with a more complex, vague, or exploratory information need. And a system that gives the user "tools" to manage that complexity, clarify that vagueness, or organize their exploration process is going to be far superior to one that just returns "www.thecausesofwar.com" or somesuch...because a tool-based system will be the one that, as Valdes-Perez says, lets the user be "dumb" by having the system do all the heavy thinking.

Pranav Dandekar said...

Here's a quote from this article on Universal Search (http://www.clickz.com/showPage.html?page=3626159) that reminded me of the Amazon joke you'd mentioned in this post:

"...Another concept that was batted around this week was what some might consider to be search's Holy Grail: a single search result. The thinking is if an engine understood you -- really, really understood you and your search query -- you'd ideally get one result to your query. It would be that one result that solved your problem."

Its funny how the same thoughts echo around the world and appear in different forms at different times.