GaGaTak

If you ask a librarian about a book on Mexico, the librarian will surely ask you to specify… Are you looking for a travel guide, history book or something completely different? Search engines could also benefit from the same approach.

With most existing online systems, a user makes an information request in a couple of words, and the search engine returns a list of documents ranked by relevance. Search technologists are busily working on relevance-ranking algorithms and question-answering systems so that they can read as much as possible into a query without asking any more of the user. But information-retrieval researchers suggest that these approaches have reached a point of diminishing returns. A search engine cannot reliably surmise the user’s intent from a single query.

What we need is human-computer information retrieval (HCIR), a term coined by University of North Carolina professor Gary Marchionini. The HCIR approach advocates for tools that bring human intelligence and attention actively into the search process. Rather than guessing what users need, these tools provide users with opportunities to clarify and elaborate their intent. If the engine isn’t sure what users want, it just asks them. (For another approach to information retrieval, see “Search Me“.)

The HCIR approach evokes what librarians call reference interviews. Indeed, HCIR leans heavily on techniques from library science, such as faceted information retrieval. Adapted for Internet use over the past decade, faceted search extends keyword search by allowing users different ways to refine queries. A search for “Mexico” might offer refinements by topic (history, demographics), language (Spanish), date published, and so on. Not surprisingly, this approach is popular for online libraries, but it has also become a staple of online shopping; Home Depot’s website is an example. HCIR transforms search engines from black-box matching engines to conversational librarians. The core technical challenge is no longer ranking the results but, rather, summarizing and organizing them so that users can interact with them. HCIR offers users the transparency, control, and guidance to establish, elaborate, and resolve their information needs.

It’s fun to work on algorithms that guess users’ intentions, and the temptation to push the limits of purely technical solutions can be irresistible. But sometimes the best approach is the most obvious one. We may do well to follow a bit of advice passed on by Nobel laureate Richard Feynman in his book Surely You’re Joking, Mr. Feynman! When discussing with a bar mate how best to pick up women, Feynman recounts, that sage soul averred, “You just ask them.”

Firefox is a open source browser and it holds almost a quarter of the entire market. A mobile version of firefox called the Fennec is going to be launched soon. Similar to Firefox, Fennec allows the developers to add different functions like bookmark synchronizing tools and even video players. More to it, designers are planning to make navigation even simpler. Take the <Awesome Bar> for example. It drops down a complete list of recently visited websites or recently searched keywords. Mozilla has released a different version for Nokia’s N810 already. Later in the year, it is planning to bring out a new version for Windows Mobile. And also versions for that run on the Symbian operating system.

Researchers are increasingly using cell phones to better understand users’ behavior and social interactions. The data collected from a phone’s GPS chip or accelerometer, for example, can reveal trends that are relevant to modeling the spread of disease, determining personal health-care needs, improving time management, and even updating social-networks. The approach, known as reality mining, has also been suggested as a way to improve targeted advertising or make cell phones smarter: a device that knows its owner is in a meeting could automatically switch its ringer off, for example.

Now a group at Dartmouth College, in Hanover, NH, has created software that uses the microphone on a cell phone to track and interpret a user’s activity. The software, called SoundSense, picks up sounds and tries to classify them into certain categories. In contrast to similar software developed previously, SoundSense can recognize completely unfamiliar sounds, and it also runs entirely on the device. SoundSense automatically classifies sounds as “voice,” “music,” or “ambient noise.” If a sound is repeated often enough or for long enough, SoundSense gives it a high “sound rank” and asks the user to confirm that it is significant and offers the option to label the sound.

The Dartmouth team focused on monitoring sound because every phone has a microphone and because accelerometers provide only limited information. “When we think about sounds, we don’t typically think that they can also represent a location that has a unique signature,” says Andrew Campbell, a professor of computer science at Dartmouth and a lead researcher on the project. The researchers made sure the program is small, so that it doesn’t use too much power. To address privacy concerns, they designed SoundSense so that information is not removed from the device for processing. Additionally, the program itself doesn’t store raw audio clips. A user can also tell the software to ignore any sounds deemed off limits.

In testing, the SoundSense software was able to correctly determine when the user was in a particular coffee shop, walking outside, brushing her teeth, cycling, and driving in the car. It also picked up the noise of an ATM machine and a fan in a particular room. The results of the experiments will be presented this week at the MobiSys 2009 conference, in Krakow, Poland.

“The SoundSense system is our first step in building a system that can learn [user behavior] on the go,” says Tanzeem Choudhury, an assistant professor at Dartmouth who was also a leader on the project and a TR35 winner. Choudhury says that enabling the software to learn to recognize new sounds will be essential for practical applications. “A system that can recognize sounds in a person’s life can be used to search for others who have the same preferences,” she says. Using sounds to classify events can give users feedback on their daily activities for health or time-management applications, she adds.

Lately, search engines have started focusing more on providing answers to specific questions. Put “capital of Botswana,” for example, into most search engines, and they’ll happily produce both the correct answer and links to relevant sites. But most search engines are of little use with more abstract queries, such as “Which book should I read?” or “What should I make for dinner?”

Hunch, a website that launches to the public today, hopes to be the answer to these questions and many more besides.

Hunch begins “where a search engine leaves off,” according to cofounder Caterina Fake, who previously cofounded the photo-sharing site Flickr and later worked on Yahoo Answers. Fake points out that a normal search engine would provide a user interested in buying a digital camera with links to hundreds of sites that review and compare the latest models. The user then has to sort through that information and figure out which camera is right for her.

In contrast, Hunch asks a few simple, multiple-choice questions, including “What type of photography are you interested in?,” “Do you want a ‘point and shoot,’ an SLR, or a Rangefinder camera?,” and “How much zoom do you want?” before recommending a specific model.

The site offers personalized recommendations for all manner of queries. Although many of the questions already on the site are lighthearted, there’s serious computer science under the hood.

After a user creates an account and logs in to Hunch, she has the opportunity to answer all manner of questions in a box labeled “Teach Hunch About You.” As the user runs through these questions, Hunch builds up reams of data to help with the recommendations that it makes.

In order to fine-tune its recommendations, Hunch balances a user’s responses to questions with information from her profile. Users can indicate whether Hunch’s recommendations were good or not, and this information will help adjust the factors that guide the site’s algorithms in the future.

Fake believes that many existing recommendation systems, such as those used by Amazon or Netflix, struggle because the data that they collect relates to a narrow range of topics. She thinks the problem is that they only have users’ book or movie ratings to work with. “Whether you like Napoleon Dynamite could have something to do with whether you played a lot of pinball as a child,” Fake says.

Fake sees Hunch as a grand experiment, but its success will depend on users’ willingness to generate new content for the site and provide feedback to train its algorithms. Although the company seeded the site with some survey questions and topics, most of what’s there now was added by users themselves during beta testing, Fake says.