Correlation Technology Demonstration Slides

A Correlation Technology Demonstration
Data

Q:

What is Correlation Technology?

A:

Correlation Technology seeks to emulate the way the human brain
acquires, stores and utilizes information

COPYRIGHT 2O1O MAKE SENCE FLORIDA, INC. ALL RIGHTS RESERVED
DISTRIBUTION WITHOUT WRITTEN PERMISSION OF MAKE SENCE FLORIDA, INC IS PROHIBITED

Please note:
The Correlation Technology Platform (CTP) is not an end user
application. The CTP is intended to support deployment of vertical
market specific software applications which incorporate Correlation
Technology. Some examples of vertical markets which have been
identified as candidates for Correlation Technology solutions
include Legal E-Discovery, Recruitment, and Market Research
Qualitative Survey Analysis. Deployment of the CTP in each
vertical market requires specialization and extension of the
Platform. Specializations include customized analysis (called
Refinement) of the Correlation result set (called the Answer
Space), and customized GUI creation appropriate to the specific
needs of each application.
The GUI and method of results presentation which follow will
almost certainly never be used by any real-world application. The
sole purpose of this presentation is provide a look “under the hood”
of Correlation Technology, giving you a first introduction to this
new, patented method of addressing data.

Query Screen with Blank Text Entry Fields

The first step is to ask the system if
the terms of interest can actually be
found in the “corpus” or collection of
documents which make up the
Correlation “domain”.

For this demonstration, the “corpus” is the entire contents of
Wikipedia

Wikipedia Transformed Into Knowledge Fragments
Correlation Technology creates “Knowledge Fragments” by
decomposing the documents in the corpus using a patented
method. Our method does not rely on syntactical sentence
decomposition, and Knowledge Fragments are not phrases.

Wikipedia has about 3.5 M documents

We exclude documents with no data, or that are only used by Wiki editors

N-Dimensional Query Terms Entered

An N-Dimensional Query asks the question,
“Identify all the connections from an Origin term or
concept to a Destination term or concept.” The
Origin and Destination must represent a lexical
and/or semantic disjunction. Although Origin
and/or Destination can be “entities” (people,
places, organizations, or things), for a standard
English language expository corpus such as
Wikipedia, other types of terms or concepts yield
more interesting results. With an average article
length of 724 words, Wikipedia is a “broad but
shallow” corpus, meaning that many topics are
represented, but not much content is present about
the average topic. Like all information retrieval
systems, Correlation Technology can not find
connections from Origins to Destinations if no data
about the Origins or Destinations is present in the
corpus. Correlation Technology can not
manufacture what is not there.

Actual Terms Selected for Correlation

“Knowledge Fragments” which match or are related to the
entered Origins are found. The “Rank” represents the number
of unique Knowledge Fragments with the same base value.
Only “checked” values will be used in the actual Correlation.
“Knowledge Fragments” which match or are related to the
entered Destinations are found.

Note: “Contexts” are not binary filters. They
expand rather than limit the answers returned,
and “guide” the Correlation process.

Although “Knowledge Fragments” which match or are related to
the entered Contexts are found, none have been chosen to be
used in this particular Correlation.

Execution Screen Parameters Set

These are the terms or concepts to be
Correlated using the Knowledge
Fragments found in the decomposed
Wikipedia corpus.
These are some governors used to
manage the Correlation process.
These governors are present for
demonstration purposes only. In
actual Correlation applications, the
systems will typically be provisioned to
run to exhaustion.

Execution Process Messages

These messages show that thousands
of correlations have been constructed
connecting Origins to Destinations.

Correlation Metrics

Correlation Technology results are not achieved by
using a reasoner iterating over a graph. There is
no graph. Every single successful correlation
starts as a trial. Starting with each Origin
Knowledge Fragment, all Knowledge Fragments in
the Correlation Technology data store are
examined for those that can be associatively linked
– first to the Origin, and then to each iteratively
linked Knowledge Fragment until either a
Destination Knowledge Fragment is found, or no
Knowledge Fragment can be associated. In the
30 seconds elapsed time permitted for this
correlation, 2.4 billion trials were performed against
the data store of 278 million Knowledge
Fragments. Only about 70 thousand correlations
were successful. These results would require the
equivalent to several lifetimes of human effort.

Execution Results Grouped Top Level

BTW: “Zanjan” is a province of Iran,
and a center of Islamic scholarship.

In order to demonstrate some aspects of
Correlation Technology, three views of the result
set of correlations will be presented. The first is a
simple “tree” hierarchical representation of the
Answer Space of correlations. Once again, please
note that this is not suggested as an end user GUI.
Rather, this is equivalent to performing a query
against a large RDBMS and scrolling through all
the rows returned in the result set. There are four
Origin Knowledge Fragments for which successful
correlations have been constructed. Each
“destination” is unique, meaning that at least one
correlation has been found (hundreds or thousands
of correlations can be constructed from each origin
to each unique destination). By expanding the
“trees”, some destinations and correlations can be
displayed.

Execution Results Grouped 1st Origin 1st 10 Destinations

For Origin Knowledge Fragment “1”, correlations
have been constructed to 897 Destinations. The
first 10 Destinations are shown. Correlation counts
are shown. All these correlations will start with
“average population density in Zanjan” and will end
with the given Destination, such as 1.1 “links to
terrorism”, or 1.2 “linked to terrorism”. Each
correlation is unique, and may be constructed from
completely different sets of Knowledge Fragments.

Execution Results Grouped 1st Origin 2nd Destination Correlations

These are correlations showing just three of the thousands of
paths linking “average population density in Zanjan” to
“terrorism”. Our patents cover dozens of ways to associate
Knowledge Fragments when constructing correlations.

Correlations can be suggestions, assertions, or statements of fact. A relation between
Origin and Destination is required and proven, but the characterization of that relation,
the relevancy of that relation, and the trusted-ness of that relation is dependent upon the
actual Correlation Technology based application. The Refinement components of the
CTP, which analyze and filter the contents of the Answer Space, are critical to applying
business rules, integrating other processes, and achieving the objectives of vertical
application specialization.

Execution Results Grouped 1st Origin 2nd Destination
2nd Correlation Resources

Although not actually required for “pure”
Correlation, each Knowledge Fragment retains a
bibliographic reference to the “resource(s)” – such
as web pages or documents – from which the
Knowledge Fragment was extracted.

When an identical Knowledge Fragment is
extracted from more than one resource, all the
“contributing resources” are captured.

Execution Results Grouped 1st Origin 2nd Ten Destinations

Some additional Destinations to which successful
correlations from Origin “1” have been constructed.
Obviously, all 897 Destinations can not be
displayed for this presentation.

Execution Results Grouped 1st Origin 50th Destination Correlations

More Correlation examples. Notice that every
possible permutation of pathway from the Origin to
the Destination is captured.

Execution Results Grouped 2nd Origin 1st 10 Destinations

first 10 Destinations are shown. All these
correlations will start with “average population
density of people” and will end with the given
Destination, such as 2.1 “state to counter
terrorism”, or 2.2 “state provide counter terrorism”.

Execution Results Grouped 2nd Origin Last 10 Destinations

As with the Origin “1”, not all Destinations can be
displayed for this presentation.

Execution Results Grouped 2nd Origin 52nd Destination Correlations

Correlations linking “average population density of
people” to “age of terrorism”.

Execution Results Grouped 2nd Origin 52nd Destination
1st Correlation Resources

Notice that *any* resource can contribute
Knowledge Fragments to a correlation.
Correlation Technology is not a documentcentric solution. Each document is subjected
to a one-way, exhaustive transformation into a
collection of Knowledge Fragments which
represent the entire “Knowledge Payload” of
the original document. Each resulting
Knowledge Fragment is an independent
knowledge object, and each is “equally
eligible” for Correlation.

Execution Results Grouped 3rd Origin 1st 10 Destinations

first 10 Destinations are shown. All these
correlations will start with “average population
density of area” and will end with the given
Destination, such as 3.1 “United States backing
state terrorism”, or 3.2 “United States attacked
Islamic terrorism”.

Execution Results Grouped 3rd Origin 32nd Destination 1st Correlation

Here is a correlation linking “average population
density of area” to “United States fight terrorism”.

Execution Results Grouped 3rd Origin 32nd Destination

Here’s the contributing resources for
Correlation 3.32.1

Execution Results Grouped 4th Origin 58th Destination Correlations

Here are Correlations linking “population density of
inhabitants” to “cities consumed terrorism”.

Execution Results Grouped 4th Origin 58th Destination

Here’s the contributing resources for
Correlation 4.58.1

Execution Results Correlation Search Engine View

From 3.5M Wikipedia documents, only 1,946 are proven to have relevance to the
complex N-Dimensional Query relating population density to terrorism.
This a sample of Refinement. By
analysis of the Answer Space (in this
case a simple frequency distribution of
contributing resources), a ranked listing
of the contributing resources most
significant to this particular NDimensional Query can be achieved.
This type of listing is generated by the
patent-pending Correlation Search
Engine (CSE). Because this Correlation
execution was not permitted to run to
exhaustion, some anomalies show up in
the list here. From the contents of the
list, however, it should be clear that a
properly provisioned full term execution
of the NDQ would accurately identify and
rank only those Wikipedia articles most
significant to the query terms.

Execution Results Flat View 221 - 230
Here is just another simple way to review
the contents of the Answer Space.

The Business of Correlation Technology
• Make Sence, Inc. controls the world-wide licensing of
Correlation Technology. Inquiries regarding licensing of
Correlation Technology for any vertical market
application should be addressed to Carl Wimmer
c.wimmer@correlationconcepts.com
• Make Sence Florida, Inc. is the US-based research and
development arm of Make Sence, Inc. and is responsible
for all technical aspects of the Correlation Technology
Platform. Address all technical inquiries to Mark Bobick
m.bobick@correlationconcepts.com
For Business Inquiries:
Contact: Carl Wimmer
carl@makesence.us
Mobile: (702) 767-7001

For Technical Inquiries:
Contact: Mark Bobick
m.bobick@correlationconcepts.com
Mobile: (702) 882-5664

Correlation Technology Demonstration Slides

Recommended

More Related Content

What's hot (19)

Viewers also liked (20)

Similar to Correlation Technology Demonstration Slides (20)

More from s0P5a41b (8)

Recently uploaded (20)

Correlation Technology Demonstration Slides