Database Reference
In-Depth Information
Gutierrez:
How would you describe your work to a layperson at a cocktail
party?
Tunkelang:
If you use a search engine, perhaps you take for granted that
the search engine knows what you mean and returns what you're looking for.
Except when it doesn't, and you curse our Skynet overlords. My team works
on converting what you say—a few words in a search box—into what you
mean and then providing the most relevant results. For example, that “ford” is
a surname in the context of “john ford”, but a company name in the context
of “ford engineer”. We do this by learning from our data. In particular, how
words are used in our search index and how people use words when they
search. We also get feedback on when we get it right, based on how people
engage or don't engage with the results we show. We use that feedback to
improve our algorithms, and hopefully we do better the next time.
Gutierrez:
What are the main types of problems being tackled in search in
social networking?
Tunkelang:
Though there are a lot of problems, you could summarize them
with three Rs—relevance, recommendations, and reputation. Relevance obvi-
ously comes up in the context of search quality, but also when we're pre-
senting feeds or streams to users and have to select interesting items from
an endless array of possibilities. Recommendations are the complement to
search. If search relevance is about giving users the information they're look-
ing for, then recommendations are about giving users the information they
need but haven't asked for. Finally, reputation encompasses a wide range of
problems, from identifying topic experts to quantifying trust.
Gutierrez:
Do you tackle all three of these problems on a frequent basis, or
do you work on a subset at any given time?
Tunkelang:
These days I focus almost exclusively on relevance. LinkedIn
has other teams that focus more on recommendations and reputation. Even
though I'm focused on relevance, I get excited about challenges in all three
areas and I do my best to contribute to our efforts in them as a company.
I also try to take part in global conversations about these topics through blog-
ging and conferences.
Gutierrez:
Who are the big thought leaders in search and social networking?
Tunkelang:
There are too many to list here, especially because working in
these fields means being an intellectual omnivore. I read research papers from
information retrieval conferences and I also look at how startups are innovat-
ing the user experience. One of the upsides of being at the epicenter of the
world's largest professional network is that new developments find their way
into my inbox through multiple channels.
