Big Changes for Search Engines
Fast, smart, personalized to suit every user's needs. And pretty. That's what the search engine of the near future will be.
Tweaking existing search engines and developing new ways to find specialized data were the subjects of two dozen papers presented at this week's 12th International World Wide Web Conference in Budapest. While search engines have improved steadily under the hood since the first days of the Web, they look and function pretty much the same as ever. But computer scientists are working on new search techniques and interfaces that could significantly alter most surfers' results pages. Expect, for example, to be able to sift through search results graphically, or to personalize Google results.
A team of researchers from Japan's Nec presented a paper titled "Mining the Peanut Gallery" describing (PDF) a tool that would allow consumers to automatically sort through product reviews. The Review Seer returns rankings of products based on user feedback culled from newsgroup and website postings. Surfers could further personalize searches by specifying what features and price ranges they are interested in.
There are still some issues to be sorted out with Seer, said Kushal Dave, a member of the team that developed the tool. One particularly pesky problem is dealing with reviews that contain a lot of negative comments, but wrap up with a final sentence exonerating the product. "Humans are so … random," grumbled one audience member.
Other speakers focused on new ways to present search results to users. Ben Shneiderman, professor of computer science at the University of Maryland in College Park, is convinced that most people are better at communicating with their computers when they can see data, rather than reading or writing it. He demonstrated several visual search tools, including the TimeSearcher, which allows seekers to see graphics illustrating search result data.
TimeSearcher allows users to define search terms on a graph, clicking and dragging to specify that results should, for example, be confined to data created or changed on specific dates. Any information that is tied to time can be searched with TimeSearcher. PhotoMesa and PhotoFinder are a combination browser and search tool specifically designed for searching digital images. Specialized search tools that work on one computer or across a private network are a step toward personalized searching of the Internet where results will be tailored to the express desires of each user.
Eventually, search engines will be able to consult cookies -- small text files placed on users' hard drives by Web servers -- and assume, from past searches, that a user is looking for one type of information and not another. The cookies could tell a search engine to return only new information, or only data targeted to the user's location. In any case, personalized searching won't be possible until Google, already known for running lightning-fast searches, is 10 times faster.
Google could increase its search speed using new techniques developed by Stanford University computer-science researchers, who presented a paper on their findings at the conference. "We've gotten a lot of press attention today from this work," said Stanford researcher Sepandar Kamvar. "Some of it is not quite correct. One major caveat: Google will not run five times faster if our research is implemented, but we do expect a 30 percent speed-up."
According to Kamvar, three simple-to-implement tweaks to Google would measurably boost search speed, "freshen" the results and -- if the research is further refined -- eventually allow for personalized searches. Google's owners haven't said whether they will adopt the Stanford research, but Google co-founder Sergey Brin was in the audience.
Source: Wired News.