Researchers classify Web searches

Apr 10, 2008

Although millions of people use Web search engines, researchers show that – by using relatively simple methods – most queries submitted can be classified into one of three categories.

Jim Jansen, assistant professor in Penn State's College of Information Sciences and Technology, worked with IST undergraduate Danielle Booth and Amanda Spink, Queensland University of Technology, to find that Web search engine users are doing primarily informational, navigational or transactional searching.

Informational searching involves looking for a specific fact or topic, navigational searching seeks to locate a specific Web site and transactional searching looks for information related to buying a particular product or service.

The research was the first published work of its kind done using actual searching data, with the aim of real-time classification. Researchers analyzed more than 1.5 million queries from hundreds of thousands of search engines users. Findings showed that about 80 percent of queries are informational and about 10 percent each are for navigational and transactional purposes.

Jansen and his colleagues arrived at those results by selecting random samples of records and analyzing query length, the order of the query in the session and the search results. These fields helped the team develop an algorithm that classified the searches with a 74-percent accuracy rate.

"Other results have classified comparatively much smaller sets of queries, usually manually," Jansen said. "This research aimed to classify queries automatically.

"Our findings have broad implications for search engines and e-commerce if they can classify the user intent of queries in real time. This is why we wanted a computational undemanding algorithm," Jansen continued. "It proves the 80/20 rule that 80 percent of the cases can be achieved with these clear-cut methods."

The paper "Determining the informational, navigational and transactional intent of Web queries" will appear in the May 2008 issue of Information Processing & Management. The article is currently available online.

The Penn State researcher said he plans to continue this research using a more complex algorithm that will hopefully yield a 90-percent accuracy rate using similar searching criteria.

Source: Penn State

Explore further: Fighting the next generation of cyberattacks

Related Stories

Hubble Source Catalog: One-stop shopping for astronomers

Mar 14, 2015

Astronomers at the Space Telescope Science Institute and the Johns Hopkins University, both in Baltimore, Maryland, have created a new master catalog of astronomical objects called the Hubble Source Catalog. ...

Build your own Siri: An open-source digital assistant

Mar 11, 2015

An open-source computing system you command with your voice like Apple's Siri is designed to spark a new generation of "intelligent personal assistants" for wearables and other devices. It could also lead to much-needed advancements ...

Recommended for you

Fighting the next generation of cyberattacks

Apr 16, 2015

The next generation of cyberattacks will be more sophisticated, more difficult to detect and more capable of wreaking untold damage on the nation's computer systems.

Algorithm able to identify online trolls

Apr 14, 2015

A trio of researchers, two from Cornell the other from Stanford has developed a computer algorithm that is capable of identifying antisocial behavior as demonstrated in website comment sections. In their ...

User comments : 2

Adjust slider to filter visible comments by rank

Display comments: newest first

Valentiinro
not rated yet Apr 10, 2008
I am wondering where porn fits into this scheme. There is quite a large amount of that on the internet. Informational I suppose?
gopher65
5 / 5 (1) Apr 11, 2008
Definitely Informational.

Please sign in to add a comment. Registration is free, and takes less than a minute. Read more

Click here to reset your password.
Sign in to get notified via email when new comments are made.