|
Susan Dumais
 The search engine is arguably the most useful thing on the Internet. Just
fill in the blank with a keyword, like "salsa," and up pops a list of web
pages. Awesome! But this miracle, unimaginable a decade ago, is not good
enough for Susan Dumais. No. She thinks search engines should be smarter,
that they should be able to tell the difference between Mexican condiments
and Afro-Caribbean music and the dance you do to that music.
Dumais is from Maine. She majored in math and psychology at Bates
College, where she was captain of the basketball team. She was a
researcher at Bellcore and at Bell Labs in New Jersey for 17 years
before she came to Microsoft. She has a great big fast new boat that she
likes to drive on Lake Washington. She knows how to program in PERL and
C++. She's very cool.
She's really enjoying herself at Microsoft Research. "I've been doing
an exciting mix of things, working with fascinating, smart people. There
are a lot of people here with complementary skills. If you have a
problem, you can just walk down the hall and get people interested in
your stuff."
Her stuff is all about human-computer interaction and it's really
important right now, as software grows more capable and more complex,
"Microsoft wants to ensure that people can actually use it."
Dumais' work on search engines, in concert with John Platt of the
Signal Processing Group and David Heckerman of the Machine Learning and
Applied Statistics Group, includes both the back end algorithms that
make them go and the face that they present to the user.
Her designs have been used to improve the search function on this
website and on the popular msn.com site. "With search on the web, you
can evolve how you present the results," she says. "That's something you
can't do with shrink-wrapped software. Web services are much more
dynamic."
The next wrinkle in search engine design is grouping results
according to category, so that if you're interested in the Afro Cuban
beat you don't have to sort through a lot of recipes. It's a program
that works with existing search engines, and she shows how it works with
a search for "salsa" on Alta Vista. Sure enough, the results are
sensibly divided into categories. Let's try it with Google. Whoops! An
error message. Turns out it was a change in the website's output format,
"which happens all the time on the web," she says.
"This is an amazingly exciting place," she says. Research's proximity
to the product teams, who are keenly interested in innovations that can
make their products better, is a powerful inducement to come to work in
the morning. "Interacting with people who care about what you do, it's
really seductive to see your ideas taken up by the product groups and
turned into products for millions of people," Dumais says. "It was just
tremendously stimulating, not just intellectually but emotionally. That
was the biggest surprise: how much fun it was."
|
