The Daedalus
A predictive search suggestion and exploration engine.

Information Exploration, LLC

IE
Welcome

Google estimates that 99% of all
searches are answered on their first
page, but at 2 million queries “per
minute” that leaves 2,880,000,000
unsatisfied customers every

© 2013 Information Exploration, LLC

day!
Welcome
“Daedalus is a user driven
solution to Big Data.”
“Daedalus is predictive search
visualization.”
“Daedalus lets you discover what
you are looking for, on the web,
at a store and in the lab.”

© 2013 Information Exploration, LLC
Outline

OUTLINE

© 2013 Information Exploration, LLC
Our Team
Sean Connolly – Human Computer Interaction

•

–
–
–

•

Brent Kievit-Kylar – Natural Language Processing
–
–
–

•

Bachelor’s Degrees from Duke University in Psychology and English.
In 2000, he sold the start-up Scriptshark.com with movie producers Roy Lee and Ed Kashiba (The
Departed, G-Force).
Currently holds a MS/MA in Human-Computer-Interaction-Design and Telecommunications and
researches and produces American 3D movies. He combines visual language with 3D interactions in the
Daedalus.

Bachelor’s Degree in Computing from Queens University.
Currently researches Natural Language Processing as a PHD candidate in Mike Jones’s computational
cognition lab.
His recent work in information visualization led to a publication in Behavior Research Methods (2012)
and the Castellan Award in the field of computational psychology for visualizing word similarities.

Rick Connolly – Strategic Advisor
–
–
–

Bachelor’s Degree Business and Finance from Wake Forest University
He was Vice president Sales and Operations of a private company with 10,000+ employees and
eventually President of a spin off from parent company.
Currently, SVP Strategic Sales and Marketing at a publicly traded 600,000+ employee global
organization.

© 2013 Information Exploration, LLC
The Problem
• Search algorithms have been improving; yet,
the interface has remained largely the same

1998

2013

© 2013 Information Exploration, LLC
The Problem
• The problem is not with the retrieval algorithms, but
with the polysemous nature of natural languages1
• Humans only generate the same words to describe
common objects 20% of the time
– This extends to search queries as well (Fidel, 1985)

1Furnas,

Landauer, Gomez, and Dumais (1987)

© 2013 Information Exploration, LLC
The Problem

McRae K et al (2005)
© 2013 Information Exploration, LLC
The Problem
The successful candidate will have an excellent academic record
with demonstrated laboratory skills and the ability to work in an
interdisciplinary team environment. Experience in
chromatography and wet analytical chemistry is desirable. The
candidate for this position should have a B.S. or M.S. or PhD
degree in Chemistry with at least 5 years of relevant experience.
Develops and validates analytical methods (HPLC and GC)
for raw materials, intermediates and drug substances.
Writes standard operating procedures (SOPs) and
qualification protocols for equipment, instrumentation and
operations in cGMP processes.

© 2013 Information Exploration, LLC
The Problem

(“PhD” OR “Ph.D.” OR “Dr.” OR “Doctor”)
OR (Master* OR “M.S.” OR “MS”) OR
(Bachelor’s OR “B.S.” OR “BS”)

© 2013 Information Exploration, LLC
The Problem
The successful candidate will have an excellent academic record
with demonstrated laboratory skills and the ability to work in an
interdisciplinary team environment. Experience in
chromatography and wet analytical chemistry is desirable. The
candidate for this position should have a B.S. or M.S. or PhD
degree in Chemistry with at least 5 years of relevant experience.
Develops and validates analytical methods (HPLC and GC)
for raw materials, intermediates and drug substances.
Writes standard operating procedures (SOPs) and
qualification protocols for equipment, instrumentation and
operations in cGMP processes.

© 2013 Information Exploration, LLC
The Problem
chem* AND ((((“PhD” OR “Ph.D.” OR “Dr.” OR
“Doctor”) OR (Master* OR “M.S.” OR “MS”) OR
(Bachelor’s OR “B.S.” OR “BS”) AND (“HPLC” OR
“H.P.L.C.” OR “High-Performance Liquid
Chromatography” OR “High-Pressure Liquid
Chromatography” OR “High Pressure LiquidChromatography” OR “High-Performance-Liquid
Chromatography” OR ““High-Pressure LiquidChromatography” ) AND (“Gas Chromatography” OR
“Gas-Chromatography” OR “GC” OR “G.C.”)) AND
(cGMP OR GMP OR “G.M.P.” OR “c.G.M.P.” OR “Good
Manufacturing Practices” OR “Good Manufacturing
Protocols” OR “current Good Manufacturing Practices”
OR “current Good Manufacturing Protocols”)))

© 2013 Information Exploration, LLC
Introduction to Solution

• The solution is not more Boolean logic
• It is letting the user speak more directly
with the algorithm
• It is putting the people back into the
search

© 2013 Information Exploration, LLC
What it Works With
The domain space
– Meta objects - Things
– Domains - Properties
– Features - Values

© 2013 Information Exploration, LLC
What it Works With
The domain space
– Meta objects - Things
– Domains - Properties
– Features - Values

© 2013 Information Exploration, LLC
How it Works
Interactive exploration loop
• Human thinks
• Computer predicts

© 2013 Information Exploration, LLC
How it Works
Visualization
– Window for each domain and for meta objects
– Values appear as bubbles in the windows
– Color indicates human or computer values

© 2013 Information Exploration, LLC
How it Works
Semantics of the search space
– Position is key
• Close to center – more relevant
• Close to each other – more similar

© 2013 Information Exploration, LLC
Mobile Technology
• Reduce Typing
– Words are suggested and may be
selected without having to type

• Natural Interface
– Sliding nodes on a surface is more
intuitive with touch
DAEDALUS

© 2013 Information Exploration, LLC

DAEDALUS
Telling a Story

Insert video of prototype here

© 2013 Information Exploration, LLC
Testimonials
• Domain expert testing
– Philosophy PhD students that we did not know asked to use the
system.
• No information on what it is or how it worked
• Able to use the tool very quickly
• Described stories about why specific words that they had not expected
(but had shown up) should be there

• Interesting remarks
– “I think you’re undervaluing this.” Ken Green – Innovate Indiana
– “This is so easy, I feel like I’ve used this before” – multiple beta testers

© 2013 Information Exploration, LLC
Market/Value
• Consumer
– Web Search
• Google ($192 billion)
• Other Web Search ($48 billion)

– E-commerce
• Amazon ($100 billion)
• eBay ($61 billion)

– Job sites

• Industrial
– Big Data technology and
services will become a
$16.9 billion dollar
industry by 20151

• CareerBuilder ($2 billion)
• LinkedIn ($9 billion)

• $ trillion market
• 1% of 1% is $50 million

© 2013 Information Exploration, LLC

• Fluid market that is
growing rapidly
Monetization

• B2B License
– Attempt to develop corporate partners
– Learn and build on each iteration with larger companies
– Flat fee or usage based

• Commission
– Make revenue through commissioned links
– Facilitates people finding products
– Take percentage of purchase

© 2013 Information Exploration, LLC
Competition

© 2013 Information Exploration, LLC
Budget
Perfect Two Client Platforms:
Integration Specialist
Graphic Designer
Brent & Sean
Network Specialist
Sub-total 1 Trip

Platform Specialists:

$60,000.00
$12,000
$12,000
$2,000
$4,000
$30,000

(3 weeks at 100/hr)
(3 weeks at 100/hr)
(5 day at location stays)
(1 week at 100/hr)

$8,0000

Perfect One Client Platforms:
Integration Specialist
Graphic Designer
Brent & Sean
Network Specialist

$30,000.00
$12,000
$12,000
$2,000
$4,000

Additional Legal:
Technology and Licenses:
iPad, Android, iPhone
iPad, Android, IPhone licenses

Time and effort Principles:
Sean
Brent
Rick

$4,000
$4,000

(1 week 100/hr)
(1 week 100/hr)

Additional Legal:

$3,000

Time and effort Principles:

iPad specialist
iPhone specialist

$12,000

$5,000

Sean
Brent
Rick

$2,000
$1,500
$500

$4,000
$4,000
$4,000

Marketing and Sales:

$5,000
$5,000
$5,000

Marketing and Sales:

$50,000

$8,0000

Total:

$5,0000

Total:

$15,000

$100,000

© 2013 Information Exploration, LLC
Timeline

• Milestones:
–
–
–
–
–
–
–
–

First web prototype
Lawyer interviews
First Big Data Prototype
Provisional Patent
First beta-tests (with local academics)
First tests (business intelligence)
Full patent
Wide release / license / sales

© 2013 Information Exploration, LLC

January 2012
March 2012
May 2012
August 2012
November 2012
May-July 2013
June 2013
October 2013
Thank you

Thank you!
Questions?

© 2013 Information Exploration, LLC
Backup Slides
© 2013 Information Exploration, LLC
The Problem
• Thesauri and deep semantic indexing
help but still remain a source of error
• Lack of visualization prevents
comprehension of the search space
• Searchers who can’t refine their
queries often abandon search
(Pirolli, 2007)

© 2013 Information Exploration, LLC
The Solution
Guiding principles
– Consistency – Confidence
– Feedback – Exploration
– Interactivity – Insight
Early Iterations

© 2013 Information Exploration, LLC

Learning Through
Visual Interaction
Papers

Related papers we have authored:
• Connolly S. (2012) Artifacts on the horizon
• Kievit-Kylar B., Jones M (2012) Visualizing multiple word
similarity measures
• Bailey R., Connolly S., Lang A (2013) Pictures enable action,
words enable thinking
• Kievit-Kylar B., Connolly S., Allen C., Jones M (in review)
Exploiting Semantic Associations to Improve Search on the Web

© 2013 Information Exploration, LLC
Google Problems
• Examples of difficult Google searches
– Value of the company Mathematica
– Chi visualization
– Tip of the tongue – what is that word? The thing that
does the thing?
– A Google a Day
• http://agoogleaday.com/
• Game created by Google that proposes hard questions to
solve
• Can be solved with Google, but difficult

© 2013 Information Exploration, LLC
Provisional Patent
•Why we got the provisional patent
• Can’t get a full patent? No.
• It protects us from the big players
• We can keep our idea secret from other companies
while we get started

© 2013 Information Exploration, LLC
Extra Materials
© 2013 Information Exploration, LLC
Handout
Front

“Daedalus is predictive
search visualization.”
“Daedalus is a user
driven solution to
Big Data.”

The Daedalus
A predictive search suggestion and exploration engine.
search.daedalus@gmail.com
1-310-801-7642
“Daedalus lets you
discover what you are
looking for, on the web, at
a store and in the lab.”
Handout
Back
Google estimates that 99% of all
searches are answered on their first page,
but at 2 million queries “per minute”
that leaves 2,880,000,000
unsatisfied customers every

day!

This is our untapped market.
Features of Dog
The Daedalus

The Daedalus

The Daedalus

A predictive search suggestion and exploration engine.

A predictive search suggestion and exploration engine.

A predictive search suggestion and exploration engine.

The Daedalus

The Daedalus

The Daedalus

A predictive search suggestion and exploration engine.

A predictive search suggestion and exploration engine.

A predictive search suggestion and exploration engine.

Semantic Solutions from Information Exploration.pptx

  • 1.
    The Daedalus A predictivesearch suggestion and exploration engine. Information Exploration, LLC IE
  • 2.
    Welcome Google estimates that99% of all searches are answered on their first page, but at 2 million queries “per minute” that leaves 2,880,000,000 unsatisfied customers every © 2013 Information Exploration, LLC day!
  • 3.
    Welcome “Daedalus is auser driven solution to Big Data.” “Daedalus is predictive search visualization.” “Daedalus lets you discover what you are looking for, on the web, at a store and in the lab.” © 2013 Information Exploration, LLC
  • 4.
  • 5.
    Our Team Sean Connolly– Human Computer Interaction • – – – • Brent Kievit-Kylar – Natural Language Processing – – – • Bachelor’s Degrees from Duke University in Psychology and English. In 2000, he sold the start-up Scriptshark.com with movie producers Roy Lee and Ed Kashiba (The Departed, G-Force). Currently holds a MS/MA in Human-Computer-Interaction-Design and Telecommunications and researches and produces American 3D movies. He combines visual language with 3D interactions in the Daedalus. Bachelor’s Degree in Computing from Queens University. Currently researches Natural Language Processing as a PHD candidate in Mike Jones’s computational cognition lab. His recent work in information visualization led to a publication in Behavior Research Methods (2012) and the Castellan Award in the field of computational psychology for visualizing word similarities. Rick Connolly – Strategic Advisor – – – Bachelor’s Degree Business and Finance from Wake Forest University He was Vice president Sales and Operations of a private company with 10,000+ employees and eventually President of a spin off from parent company. Currently, SVP Strategic Sales and Marketing at a publicly traded 600,000+ employee global organization. © 2013 Information Exploration, LLC
  • 6.
    The Problem • Searchalgorithms have been improving; yet, the interface has remained largely the same 1998 2013 © 2013 Information Exploration, LLC
  • 7.
    The Problem • Theproblem is not with the retrieval algorithms, but with the polysemous nature of natural languages1 • Humans only generate the same words to describe common objects 20% of the time – This extends to search queries as well (Fidel, 1985) 1Furnas, Landauer, Gomez, and Dumais (1987) © 2013 Information Exploration, LLC
  • 8.
    The Problem McRae Ket al (2005) © 2013 Information Exploration, LLC
  • 9.
    The Problem The successfulcandidate will have an excellent academic record with demonstrated laboratory skills and the ability to work in an interdisciplinary team environment. Experience in chromatography and wet analytical chemistry is desirable. The candidate for this position should have a B.S. or M.S. or PhD degree in Chemistry with at least 5 years of relevant experience. Develops and validates analytical methods (HPLC and GC) for raw materials, intermediates and drug substances. Writes standard operating procedures (SOPs) and qualification protocols for equipment, instrumentation and operations in cGMP processes. © 2013 Information Exploration, LLC
  • 10.
    The Problem (“PhD” OR“Ph.D.” OR “Dr.” OR “Doctor”) OR (Master* OR “M.S.” OR “MS”) OR (Bachelor’s OR “B.S.” OR “BS”) © 2013 Information Exploration, LLC
  • 11.
    The Problem The successfulcandidate will have an excellent academic record with demonstrated laboratory skills and the ability to work in an interdisciplinary team environment. Experience in chromatography and wet analytical chemistry is desirable. The candidate for this position should have a B.S. or M.S. or PhD degree in Chemistry with at least 5 years of relevant experience. Develops and validates analytical methods (HPLC and GC) for raw materials, intermediates and drug substances. Writes standard operating procedures (SOPs) and qualification protocols for equipment, instrumentation and operations in cGMP processes. © 2013 Information Exploration, LLC
  • 12.
    The Problem chem* AND((((“PhD” OR “Ph.D.” OR “Dr.” OR “Doctor”) OR (Master* OR “M.S.” OR “MS”) OR (Bachelor’s OR “B.S.” OR “BS”) AND (“HPLC” OR “H.P.L.C.” OR “High-Performance Liquid Chromatography” OR “High-Pressure Liquid Chromatography” OR “High Pressure LiquidChromatography” OR “High-Performance-Liquid Chromatography” OR ““High-Pressure LiquidChromatography” ) AND (“Gas Chromatography” OR “Gas-Chromatography” OR “GC” OR “G.C.”)) AND (cGMP OR GMP OR “G.M.P.” OR “c.G.M.P.” OR “Good Manufacturing Practices” OR “Good Manufacturing Protocols” OR “current Good Manufacturing Practices” OR “current Good Manufacturing Protocols”))) © 2013 Information Exploration, LLC
  • 13.
    Introduction to Solution •The solution is not more Boolean logic • It is letting the user speak more directly with the algorithm • It is putting the people back into the search © 2013 Information Exploration, LLC
  • 14.
    What it WorksWith The domain space – Meta objects - Things – Domains - Properties – Features - Values © 2013 Information Exploration, LLC
  • 15.
    What it WorksWith The domain space – Meta objects - Things – Domains - Properties – Features - Values © 2013 Information Exploration, LLC
  • 16.
    How it Works Interactiveexploration loop • Human thinks • Computer predicts © 2013 Information Exploration, LLC
  • 17.
    How it Works Visualization –Window for each domain and for meta objects – Values appear as bubbles in the windows – Color indicates human or computer values © 2013 Information Exploration, LLC
  • 18.
    How it Works Semanticsof the search space – Position is key • Close to center – more relevant • Close to each other – more similar © 2013 Information Exploration, LLC
  • 19.
    Mobile Technology • ReduceTyping – Words are suggested and may be selected without having to type • Natural Interface – Sliding nodes on a surface is more intuitive with touch DAEDALUS © 2013 Information Exploration, LLC DAEDALUS
  • 20.
    Telling a Story Insertvideo of prototype here © 2013 Information Exploration, LLC
  • 21.
    Testimonials • Domain experttesting – Philosophy PhD students that we did not know asked to use the system. • No information on what it is or how it worked • Able to use the tool very quickly • Described stories about why specific words that they had not expected (but had shown up) should be there • Interesting remarks – “I think you’re undervaluing this.” Ken Green – Innovate Indiana – “This is so easy, I feel like I’ve used this before” – multiple beta testers © 2013 Information Exploration, LLC
  • 22.
    Market/Value • Consumer – WebSearch • Google ($192 billion) • Other Web Search ($48 billion) – E-commerce • Amazon ($100 billion) • eBay ($61 billion) – Job sites • Industrial – Big Data technology and services will become a $16.9 billion dollar industry by 20151 • CareerBuilder ($2 billion) • LinkedIn ($9 billion) • $ trillion market • 1% of 1% is $50 million © 2013 Information Exploration, LLC • Fluid market that is growing rapidly
  • 23.
    Monetization • B2B License –Attempt to develop corporate partners – Learn and build on each iteration with larger companies – Flat fee or usage based • Commission – Make revenue through commissioned links – Facilitates people finding products – Take percentage of purchase © 2013 Information Exploration, LLC
  • 24.
  • 25.
    Budget Perfect Two ClientPlatforms: Integration Specialist Graphic Designer Brent & Sean Network Specialist Sub-total 1 Trip Platform Specialists: $60,000.00 $12,000 $12,000 $2,000 $4,000 $30,000 (3 weeks at 100/hr) (3 weeks at 100/hr) (5 day at location stays) (1 week at 100/hr) $8,0000 Perfect One Client Platforms: Integration Specialist Graphic Designer Brent & Sean Network Specialist $30,000.00 $12,000 $12,000 $2,000 $4,000 Additional Legal: Technology and Licenses: iPad, Android, iPhone iPad, Android, IPhone licenses Time and effort Principles: Sean Brent Rick $4,000 $4,000 (1 week 100/hr) (1 week 100/hr) Additional Legal: $3,000 Time and effort Principles: iPad specialist iPhone specialist $12,000 $5,000 Sean Brent Rick $2,000 $1,500 $500 $4,000 $4,000 $4,000 Marketing and Sales: $5,000 $5,000 $5,000 Marketing and Sales: $50,000 $8,0000 Total: $5,0000 Total: $15,000 $100,000 © 2013 Information Exploration, LLC
  • 26.
    Timeline • Milestones: – – – – – – – – First webprototype Lawyer interviews First Big Data Prototype Provisional Patent First beta-tests (with local academics) First tests (business intelligence) Full patent Wide release / license / sales © 2013 Information Exploration, LLC January 2012 March 2012 May 2012 August 2012 November 2012 May-July 2013 June 2013 October 2013
  • 27.
    Thank you Thank you! Questions? ©2013 Information Exploration, LLC
  • 28.
    Backup Slides © 2013Information Exploration, LLC
  • 29.
    The Problem • Thesauriand deep semantic indexing help but still remain a source of error • Lack of visualization prevents comprehension of the search space • Searchers who can’t refine their queries often abandon search (Pirolli, 2007) © 2013 Information Exploration, LLC
  • 30.
    The Solution Guiding principles –Consistency – Confidence – Feedback – Exploration – Interactivity – Insight Early Iterations © 2013 Information Exploration, LLC Learning Through Visual Interaction
  • 31.
    Papers Related papers wehave authored: • Connolly S. (2012) Artifacts on the horizon • Kievit-Kylar B., Jones M (2012) Visualizing multiple word similarity measures • Bailey R., Connolly S., Lang A (2013) Pictures enable action, words enable thinking • Kievit-Kylar B., Connolly S., Allen C., Jones M (in review) Exploiting Semantic Associations to Improve Search on the Web © 2013 Information Exploration, LLC
  • 32.
    Google Problems • Examplesof difficult Google searches – Value of the company Mathematica – Chi visualization – Tip of the tongue – what is that word? The thing that does the thing? – A Google a Day • http://agoogleaday.com/ • Game created by Google that proposes hard questions to solve • Can be solved with Google, but difficult © 2013 Information Exploration, LLC
  • 33.
    Provisional Patent •Why wegot the provisional patent • Can’t get a full patent? No. • It protects us from the big players • We can keep our idea secret from other companies while we get started © 2013 Information Exploration, LLC
  • 34.
    Extra Materials © 2013Information Exploration, LLC
  • 35.
    Handout Front “Daedalus is predictive searchvisualization.” “Daedalus is a user driven solution to Big Data.” The Daedalus A predictive search suggestion and exploration engine. search.daedalus@gmail.com 1-310-801-7642 “Daedalus lets you discover what you are looking for, on the web, at a store and in the lab.”
  • 36.
    Handout Back Google estimates that99% of all searches are answered on their first page, but at 2 million queries “per minute” that leaves 2,880,000,000 unsatisfied customers every day! This is our untapped market.
  • 37.
    Features of Dog TheDaedalus The Daedalus The Daedalus A predictive search suggestion and exploration engine. A predictive search suggestion and exploration engine. A predictive search suggestion and exploration engine. The Daedalus The Daedalus The Daedalus A predictive search suggestion and exploration engine. A predictive search suggestion and exploration engine. A predictive search suggestion and exploration engine.