38
Recruiting Solutions Social Search in a Professional Context Daniel Tunkelang LinkedIn, Head of Query Understanding 1 Daniel Workshop on Data-driven User Behavioral Modeling and Mining from Social Media

Social Search in a Professional Context

Embed Size (px)

DESCRIPTION

Keynote at CIKM 2013 Workshop on Data-driven User Behavioral Modelling and Mining from Social Media Social Search in a Professional Context Daniel Tunkelang (LinkedIn) Social networks bring a new dimension to search. Instead of looking for web pages or text documents, LinkedIn members search a world of entities connected by a rich graph of relationships. Search is a fundamental part of the LinkedIn ecosystem, as it helps our members find and be found. Unlike most search applications, LinkedIn's search experience is highly personalized: two LinkedIn members performing the same search query are likely to see completely different results. Delivering the right results to the right person depends on our ability to leverage our each member's unique professional identity and network. In this talk, I'll describe the kinds of search behavior we see on LinkedIn, and some of the approaches we've taken to help our members address their information needs.

Citation preview

Page 1: Social Search in a Professional Context

Recruiting Solutions Recruiting Solutions Recruiting Solutions

Social Search in a Professional Context

Daniel Tunkelang LinkedIn, Head of Query Understanding

1

Daniel

Workshop on Data-driven User Behavioral Modeling and Mining from Social Media

Page 2: Social Search in a Professional Context

LinkedIn connects talent to opportunity.

2

Search enables the participants in the economic graph to find and be found.

Page 3: Social Search in a Professional Context

Overview

Why do people search in a professional context?

How do we help people search in

a professional context?

Next play?

3

Page 4: Social Search in a Professional Context

4

Page 5: Social Search in a Professional Context

Scenario 1: Pleased to meet you!

5

Page 6: Social Search in a Professional Context

People search isn’t the same as web search.

6

Page 7: Social Search in a Professional Context

LinkedIn works hard to make it effortless.

7

Page 8: Social Search in a Professional Context

Even harder to reduce user effort to a few chars.

8

Page 9: Social Search in a Professional Context

Searchers use what they know to find people.

9

Page 10: Social Search in a Professional Context

Not all navigational queries are name searches.

10

Page 11: Social Search in a Professional Context

Scenario 2: Looking for new opportunities.

11

Page 12: Social Search in a Professional Context

Lots of jobs in DC for data scientists.

12

Page 13: Social Search in a Professional Context

My connections can help me get a $100k+ job.

13

Page 14: Social Search in a Professional Context

Apply, contact the recruiter, or seek a referral.

14

Page 15: Social Search in a Professional Context

Scenario 3: I know what I want when I see it.

15

Page 16: Social Search in a Professional Context

Another year, another CIKM industry event.

16

Page 17: Social Search in a Professional Context

We’ll need student volunteers, too.

17

Page 18: Social Search in a Professional Context

And some sponsors!

18

Page 19: Social Search in a Professional Context

19

Page 20: Social Search in a Professional Context

LinkedIn’s focus: entity-oriented search.

20

Company

Employees

Jobs

Name Search

Page 21: Social Search in a Professional Context

Query tagging: key to query understanding.

§  Using human judgments to evaluate tag precision. –  Extremely accurate (> 99%) for identifying person names. –  Harder to distinguish company vs. title vs. skill (e.g., oracle dba).

§  Comparing CTR for tag matches vs. non-matches. –  Difference can be large enough to suggest filtering vs. ranking:

21

Page 22: Social Search in a Professional Context

Query Tagging: An Example

22

Page 23: Social Search in a Professional Context

Detecting navigational vs. exploratory queries.

Pre-retrieval §  Sequence of query tags.

Post-retrieval §  Distribution of scores / features.

23

Click behavior §  Title searches >50x more

likely to get 2+ clicks than name searches.

Page 24: Social Search in a Professional Context

Navigation vs. Exploration: Behavior Patterns

§  Exploratory searches leads to ~5x more clicks per search than navigational searches.

§  Clicks on 2nd-degree connection more than 2x as likely to lead to invitation from exploratory vs. navigational search.

§  For navigational queries, 1st degree > 2nd degree > …

§  For exploratory queries, 2nd and 3rd degree > 1st degree.

24

Page 25: Social Search in a Professional Context

Query expansion for exploratory queries.

25

software patent lawyer

Query expansions derived from reformulations.

e.g., lawyer -> attorney

Page 26: Social Search in a Professional Context

LinkedIn search is personalized.

26

kevin scott

Page 27: Social Search in a Professional Context

But global factors matter.

27

Page 28: Social Search in a Professional Context

Relevant results can be in or out of network.

28

§  Searcher’s network matters for relevance. –  Within network results have higher CTR.

§  But the network is not enough. –  About two thirds of search clicks come from out of

network results.

Page 29: Social Search in a Professional Context

Personalized machine-learned ranking.

29

§  Data point is a triple (searcher, query, document). –  Searcher features are important!

§  Labels: Is this document relevant to the query and the user? –  Depends on the user’s network, location, etc. –  Too much to ask random person to judge.

§  Training data has to be collected from search logs.

Page 30: Social Search in a Professional Context

How to train your model.

30

§  Train simple models to resemble complex ones. –  Build Additive Groves model [Sorokina et al, ECML ’07],

which is good at detecting interactions. §  Build tree with logistic regression leaves.

§  By restricting tree to user and query features, only regression model evaluated for each document.

β0 +β1T (x1)+...+βn xn

α0 +α1P(x1)+...+αnQ(xn )

X2=?

X10< 0.1234 ?

γ0 +γ1R(x1)+...+γnQ(xn )

Page 31: Social Search in a Professional Context

31

Page 32: Social Search in a Professional Context

32

Page 33: Social Search in a Professional Context

Make search truly entity-centric.

33

results

results

Page 34: Social Search in a Professional Context

Use the search box to surface task intent.

34

I am…

looking for a job… at LinkedIn in Fiji trying to hire…

software engineers web developers

interested in learning about… Hadoop NoSQL

Page 35: Social Search in a Professional Context

It takes two to connect talent to opportunity.

35

Page 36: Social Search in a Professional Context

LinkedIn: connecting talent to opportunity.

36

Search: enabling the participants in the economic graph to find and be found.

Page 37: Social Search in a Professional Context

Thank you!

37

238,

Page 38: Social Search in a Professional Context

Want to learn more?

§  Check out http://data.linkedin.com/search.

§  Contact me: [email protected]

http://linkedin.com/in/dtunkelang

§  Did I mention that we’re hiring? J

38