Name: Deep Semantic vs. Keyword and Shallow Linguistic: A New Approach for Supporting Exploitation

Text: Expert System
Deep Semantic vs. Keyword and Shallow Linguistic:
A New Approach for Supporting Exploitation
Rita Joseph
Federal Government Operations
Expert System

Who we are
Expert System is the largest, fastest growing
semantic software company in the world.
We develop technology, applications and
solutions to extract, understand and share
information more effectively.

2

Established market presence
•  Expert System was established in Modena, Italy by three
young programmers with an idea. A few months later,
Expert System’s software was integrated into the Microsoft
Office suite.
•  Private and Profitable with Revenue doubled in the last
three years to over $15 million in 2010 and EBITDA above
20%.
•  30% of resources devoted to R&D and over $14 million
invested in the last 3 years, with $5M more planned for the
next 2 years.
•  More than 100 employees and offices in Italy, London,
Washington, D.C. and Chicago.

3

Recognized for mature and proven technology
Identified among the world’s leading information
access technology developers.
Selected one of the “Innovative Information
Access Companies Under $100M to Watch.”
Recognized for text analytics and superior
SharePoint integration capabilities.
One of the few non-Microsoft technologies in the
MS Office suite.

4

A flood of unstructured data & information
More than 80% of the knowledge on which our daily jobs
are based is unstructured (emails, documents, web
pages, articles, information from social media, etc.).
•  Over 294 Billion

emails sent daily.
•  Over 6.1 Trillion
text messages sent
in 2010.
• And what about
phone calls, faxes,
chat sessions, etc. ?
Sources: Radicati Group, ITU.

The limits of traditional approaches
Keyword Technology
or Statistics
Breaks text into single words
without considering the
context, like reading a
language that we don’t
understand:
Az IBM szokásosan nagy hangsúlyt
helyez a továbbképzésre, így
munkatársai évente számos szakmai
tanfolyamon vesznek részt.

Shallow Linguistic
Technology
Recognizes words and identifies
their most basic forms
(lemmas), but cannot
distinguish between different
meanings.
Sell -> Selling -> Sold

Neither understands the meaning of words.

6

Where semantic technology excels

One keyword, many
different meanings.
Over 231 million
results
for a single query.

7

The information we need is harder to find

Productivity of search

The increasing amount of information
• 15 Petabytes of new information a day
• 15 million searches a month
The diminishing effectiveness of search
• 1/3 of searches do not find intended results
• Over two hours a day are spent searching for information

Web
Desktop

PC Era

Social Web

Semantic
Web

Natural Language Search
Tagging

Keyword Search (Google)
Directories

Files & Folders

Databases

Amount of information

8

Why we are different
Semantic technology understands the meaning of
words in the same way you learned to read.
•  It understands the relationships
between words.
Luke (subject) has eaten (verb)
a chicken (object).
•  It understands the meaning of
words.
To eat (chicken); to consume
(oil); to destroy (sweater); to
spend (money); to rust (the
tower), etc.

9

Next generation technology

The problem of text analysis
Same word,
different
meanings

Different words,
but the same
meanings

Different words,
related
meanings

Jaguar (animal)
Jaguar (car)

Disability Legislation
Equal Opportunity Law

Organization à Company
Organization à Charity
Organization à Trade Union

11

How Cogito works

12

What is a semantic network?
A rich map of associations and meanings of words.
•  Includes all definitions of all words.
•  Includes relationships between all words.
The quality of results is derived from the richness and complexity of
the semantic network.

COGITO® English
Semantic Network:



350,000 words
2.8M relationships

13

The semantic net, the heart of Cogito
Traditional technologies can only guess the meaning of words
using keywords, shallow linguistics and statistics.
Instead, semantic networks can identify:
Terms
Abbrev.

“San Jose is an
American city.”

Concepts
Connections

Phrases

Meanings
Domains

“San Jose is a
geographic part
of California.”

14

Technology stack
Development
Studio
90% Precision

1. Morphology

Linguistic
Query
Engine

2. Grammatical

80% Precision

3. Logic
4. Disambiguation

Semantic
Network

Semantic
Semantic
Network
English
Network
Arabic

Develop and Add Custom Rules

Superior technology, tools
and customization
services maximize the
quality and the
performance of the
solution.

Italian
German
Other Middle Eastern

15

The objectives of IT

All areas
where
semantic
technology
plays a
critical role.

Source: AMR Research

16

How Expert System is unique
The Cogito semantic platform improves the
quality of results, and excels in:


Recall. Retrieves more

Productivity of search

relevant information through
search.


Precision. Retrieves a high
level of accurate results that are
relevant to your query.

• Speed. Finds information
quickly.

Amount of Information

17

Contact us

Thank You!

Rita Joseph
rjoseph@expertsystem.net
www.expertsystem.net

18

Expert System in the news

19

Document Path: ["69-201110-iss-iad-t5-expertsystem.pdf"]

e-Highlighter

Click to send permalink to address bar, or right-click to copy permalink.

Un-highlight all Un-highlight selectionu Highlight selectionh