Navigating Semantic Search

Navigating Semantic Search

Presented by:

Irina Shamaeva

Partner, Brain Gain Recruiting

May 13, 2010

The Webinar Will Begin Shortly

If you cannot hear music, or the presenter to today's webinar, please use our toll-free call in number.

Number: 888-469-1348 Pass code: 2940000

2

Agenda for Today’s Webinar:

What is Semantic Search?

Boolean vs. Semantic

Semantic Search Aspectso Word Proximity

o Abbreviations

o Synonyms

o Keyword Clouds

Semantic Featureso Web Search Engines

o Social Media

Semantic Search Engines

Semantic Search Tools for Recruiters

Resources

Q&A

3

What is Semantic Search?

Semantic is meaning

In recruiting, semantic search would ideally identify the right candidates

Job Description

ResumeResume

ResumeResume

4

Semantic Search:Areas of Implementation

These factors affect feasibility of semantic search (keep in mind when comparing systems):

o All world wide web vs. structured databases

o Narrow topic vs. “everything”

o A task to solve vs. exploration

5

What is Boolean?

Boolean means AND, OR, NOT

Boolean search syntax is the Boolean logic combined with operators, special characters, and options:

6

What is Boolean?

Note: Boolean syntax is different for different search engines and databases

7

Boolean “vs.” Semantic

Why compare the two?o Boolean is AND, OR, NOT

o Semantic is meaning

Elements of semantic search are present in existing “Boolean” search engines

New “semantic” tools often include Boolean logic

8

Sourcing Process

9

“Boolean” Approach Example

Pick keywords from a job description and create Boolean strings

Software Engineer. You’ll help build next-generation security products, working as part of an energetic team and hacking in a Linux/Mac/open source environment

The ideal candidate has: * 3+ years software development experience (Demonstrable fluency in C/C++)

* Demonstrable fluency in Ruby, or another interpreted, open source language (Python, Perl, PHP, etc.)

* A Bachelor’s degree in Computer Science, Mathematics or equivalent work experience

* TCP/IP experience, up and down the stack

* RDBMS familiarity, including entity design and SQL query troubleshooting

10

“Boolean” Approach

"software engineer” Linux "open source" C++ Ruby TCP/IP SQL > “7,000+" results (an estimate); not right – mostly job posts

"software engineer” Linux "open source" C++ Ruby TCP/IP SQL -jobs -job> about 700+ results (not bad!; many are resumes, some are not)

intitle:resume | inurl:resume "software engineer” Linux "open source" C++ Ruby TCP/IP SQL > about 75 results (good but we want more)

intitle:resume | inurl:resume "software engineer” | developer Linux | UNIX "open source" C++ Ruby | Perl | python | PHP TCP/IP SQL> about 700 results; pretty good; can play with it and look at the results

11

“Semantic” Approach May…

Recognize the title, keywords, must-haves, locations

Suggest alternative terminology, abbreviations, target companies, colleges

Search for resumes that have target titles, must-have keywords in the recent experience

12

Still, Any Tool Requires Human Input

Real life challenge:

“Тhey would like the functional candidates to come from similar consulting firms or recently in a consulting firm. They want to see some stability. For the technical roles – they need consulting, not out of industry. Except for the hard-to-

find roles like Business objects.”

13

Semantic Search Angles

Word Proximityo Example: “managed NEAR people”

Abbreviationso PwC = Pricewaterhousecoopers; “Big 4” = “Big Four”

Synonymso Software Engineer = Developer SYNONYMS

ARE ALL THE SAME

TO ME

SYNONYMSARE ALL

THE SAMETO ME

14

More Semantic Search Angles

Weighted wordso Example: titles vs. must-haves vs. nice-to-

haves in resumes

Keyword cloudso Relevant keywords may be added to search

Rankingo Semantic search may order results based on

content, not on popularity

15

Semantic Features – Web Search Engines

Auto-stemming (child children)

Selected recognition of special characters:

o C++, C#

Some results will be relevant but will not have your keyword

16

Semantic Features – Web Search Engines

Google offers similar search strings

Google personalizes results, making them more relevant over time

17

Semantic Features – Web Search Engines

Asterisk * (almost) allows for proximity search

18

Semantic Features – Web Search Engines

Tilde ~ allows to search for synonyms

19

Semantic Features – Social MediaStructured profiles allow meaningful search (but no synonyms recognition)

20

Semantic Features – Social Media

Extra search capabilities:o Tag search: #hashtags on Twitter; tagged blog posts

o Search within targeted groups and communities: LinkedIn groups

21

Need For Improved Search

The web search lacks the capacity to match our needs as recruiters. There’s no sure way to:

o Find resumes among documents

o Find people in the right locations

o Find skills matching those in a job description

o Not to miss resumes with similar terms

We need to review lots and lots of resumes trying to find the best

22

Need for Improved Search

Goal: o Spend less time searching and more time

talking to the right candidates

Possible solutions for recruiters:o Parsing toolso Semantic search

23

Parsing Tools

By parsing search results we can narrow them down to those that are likely to be on target

Example: Find resumes using the keyword Microsoft, parse, and pick those where Microsoft means employer

24

Semantic Search EnginesProgramming Considerations

It is easier to implement semantic search:

o For a particular task (such as searching resumes in a specific industry) than as a general tool

o For a limited set of structured documents (such as a resume database), than for the web

25

General Semantic Search Engines… …may be useful for research…

26

…but won’t solve recruiting problemsScreenshot from a semantic search engine

27

Semantic Search For Recruiters

There is no one “ideal” semantic tool recruiters, yetNew tools offer:o Elements of semantic search for the webo Implementations of semantic search for sets of resumes or

profileso Friendlier UI than the Boolean syntax

28

Semantic Search For Recruiters:Functionality

Help building queries

o Custom search engines

o String suggestions (restrictive)

o Use clouds of keywords for an industry

o Use lists (like “top schools”, “list of majors”, etc.)

29

Semantic Search For Recruiters:Functionality (cont.)

Allow users to give weights to keywords

Search for matches in user’s networks

Add extra semantic search capacity to existing resume databases

30

Semantic Search For Recruiters:User Interface

Boolean search strings may look cryptic but they allow control over search

A substitute UI needs to be friendly without affecting the quality of results

Look for the right tool

Incorporating Semantic Search Into Your Sourcing Strategy

Assess semantic search toolso Understand the technology behind ito Have the right expectationso Try it out for your purposes (industry, location, etc.)

Extra points if it can “learn” or be tunedo See how it integrates with other tools you useo Check whether the UI is clear and friendly

You may want to use several sourcing tools if they complement each other

31

Summary

Use semantic search capacity of search engines and Social Networks

Use general semantic search engines for research

Review new semantic search tools for recruiters – start incorporating them into your recruiting strategy

Expect increasing industry focus, adoption, and innovation in semantic recruiting in the months to come

32

33

Boolean Strings Network and Group

> http://booleanstrings.ning.com/

> LinkedIn group: http://www.linkedin.com/groups?gid=1176637

My Blog

> http://booleanstrings.wordpress.com/

I offer:

> Training DVDs and Webinars

> Sourcing Help

> Please email me at [email protected]

Thank You & Q&A