Database Reference
In-Depth Information
perform after those clicks—such as inviting a person to connect or applying
for a job. We also use human judgments—but that can only go so far when
we're evaluating a highly personalized search experience.
Gutierrez:
How does LinkedIn personalize the search experience?
Tunkelang:
The personalization of your search experience depends mostly
on information in your profile, like your company, location, industry, and your
network. For some classes of searches, network distance is a particularly
strong feature—for example, it helps us disambiguate name searches.
Gutierrez:
Search is something you have done in different roles and companies.
What about it excited you at the beginning and what excites you now?
Tunkelang:
Search is an amazing problem that just keeps giving. It seems so
easy when you first encounter it, as it probably seemed to information
scientists in the 1950s who had to invent the concept of “relevance” when
they realized indexing documents by their words wasn't enough to make them
findable. Search is the problem at the heart of the information economy. The
information is out there, if only we can find it. What's also great about search
is that it's an area full of open problems, many of them pretty fundamental.
Maybe search will be boring fifty years from now, but I doubt it.
Gutierrez:
If someone wants to get started in search today, what should
they do?
Tunkelang:
They should start by taking a class on information retrieval or
learn from the vast array of resources available offline and online. Given the
open source technology for search, they should learn by doing—for instance,
implementing a basic search engine for a public data collection. It's not hard
to get started with search.
Where things get interesting is in the details. Ranking is still an open area of
research, especially for personalized and social search applications. And there
are even more opportunities to experiment with new search user interfaces.
I'm biased, but that's where I'd direct people interested in pursuing the future
of search.
Gutierrez:
What did your work on search at Google entail?
Tunkelang:
I led at a Google team that mapped businesses in Google's local
search index—which is now part of Google+—to their official home pages. In
addition to its web search index, Google has a local search index intended to
help people find local businesses. Think of it as a 21st-century yellow pages.
My team's goal was to improve local search quality. We approached this goal
by mapping the local search index to the web search index in order to deter-
mine when local business had official home pages, and associating the two.
