Database Reference
In-Depth Information
We've also embraced more sophisticated ranking approaches. And we're
increasingly looking at two-sided relevance approaches that reflect the talent
marketplace we're trying to optimize.
Gutierrez: What do you mean by “standardized entities” and why is “things
not strings” the ultimate goal?
Tunkelang: At LinkedIn, our entities are people, companies, universities,
and similar things. Of course, people enter strings into the search box just
like with any other search engine. But they are mostly trying to refer to the
entities that participate in LinkedIn's ecosystem—or, as we like to call it, the
“economic graph.”
The “things-not-strings” idea is simple. Because the strings are intended to
refer to entities, let's remove the barrier of uncertainty between the strings
and the entities. Doing so helps enable a world of richer queries, where we
bring in relationships among entities—for example, acquired companies,
similar job titles—and provide a more structured, guided exploration of the
information space. Google and Microsoft are starting to do this for the web with
their “knowledge graph” projects, but we have an extraordinary ability to provide
this experience today because of the rich structure inherent in our data.
Gutierrez: What is a two-sided relevance search approach? And how do you
optimize it?
Tunkelang: LinkedIn is a marketplace connecting talent to opportunity at
massive scale. And a successful connection requires satisfying both of the
parties being connected. It's not enough to help a job-seeker find her dream
job or to help a recruiter find his dream candidate— the job or candidate has
to be attainable. What we end up with was a multiple-objective optimization
problem that incorporates both the searcher's desires and the likelihood of a
successful outcome.
Gutierrez: What do LinkedIn members search for?
Tunkelang: LinkedIn members search for people, jobs, companies, and
various kinds of content. For certain kinds of professionals, like recruiters
and salespeople, finding through LinkedIn is a core part of their livelihood.
For others who are less outbound as professionals, being found means having
access to unexpected career opportunities. Getting both right allows us to
best connect talent to economic opportunity.
Gutierrez: How do you measure if a search was successful or not?
Tunkelang: We have several measures that we use to evaluate search success.
We look at whether people click on results, and even more at actions they
 
Search WWH ::




Custom Search