Introducing Info Arab 's Technologies in the fields of Semantic Crawling and Semantic Analysis.

1. The information
SUPERNOVA
THE THREAT AND THE CURE
INTRODUCTION TO INFO ARAB ‘S
TECHNOLOGIES
BY; ALAA S. AL-AGAMAWI.
SEPTEMBER 2014

2. Technologies impact on data sizes…
sizes in algorithmic
scale
Invention of Paper
{China 105AD}
Invention of writing
{IRAQ 5000BC}
Invention of Printing
{Europe 1400AD}
Invention of ICT
{2000AD}
5000BC 4000BC 3000BC 2000BC 1000BC 0 1000AD 2000AD
 Technology has always played the most major role in the growth of human knowledge.
 With the ICT user generated phenomena, rate of growth is extraordinary.

3. Before the
SUPERNOVA
explodes !!! WHEN A STAR GOES OUT OF CONTROL IT
EXPLODES AND EATS EVERYTHING AROUND!
IMAGINE …
THE INFORMATION SUPERNOVA.
WHEN DATA SIZE IS OUT OF CONTROL.
TO CONTROL IT … WE NEED TO UNDERSTAND IT.

4. The ICT era…

5. The social networks…
What happens in one minute in the social networks
requires over…
1500 person days to read.

6. Strongest FRIEND and Toughest ENEMY!!!
Whether you are a movie star, a politician, a governmental agency, a
broker or a brand owner, you should give adequate attention to what
people are saying about you. In most cases, traditional monitoring is
not enough and requires tens and hundreds of person days. And still
cannot cover everything thoroughly.

7. Your ONLY
choice…
2 1
3
Go
Is to utilize ICT
to know what
the people are
saying about
you and ACT
accordingly.
Collect
Organize
Act
Anticipate Analyze

8. The Winning TEAM…
Collect Organize Analyze Anticipate Act
BI
Implementation
Partner
THIS
You should
do yourself

9. A typical solution architecture…
Data Sources
Data
Connectors
Enterprise Servers world wide web
Business
Presentation
Layer
Data Gathering
Layer
Results
Layer
Significs©

10. Proudly introducing
Fetch Technologies©

11. Historical
background…
 Fetch PC© 2006.
 Fetch enterprise© 2009.
 Fetch cloud© 2012.

12. The dilemma…
Relevant Area
Irrelevant Area Irrelevant Area
Typical crawling will bring many irrelevant data. That is…
 Irrelevant information.
 And waste of computer resources.

13. Traditional crawling techniques…
v
In many cases search results goes to an irrelevant area.

14. Structured & unstructured data…

15. Dealing with all types of Sources…
Having an API… Does not have an API…

16. Only relevant data in structured format
Title
Brief
Media
Article
body
Publisher Author Date

17. Classifies with Social media interactions…

18. Never miss what you are looking for…
You will never miss any mention that you are looking for with our semantic based
search engines. Our engines are taking care of the following...
 Basic search objects in a hierarchal format (e.g. a company, its senior
managerial staff, its main brands, sub brands ...etc).
 Totally morphed (i.e. in their singular, plural, masculine, feminine, adjectives
...etc.).
 With all their Arabic/English synonyms and antonyms.
 Taking care of spelling mistakes and grammatical errors.
 Taking care of object types and their components.

19. And everything is easily available…
RSS
XML
JSON
SaaS
Timely
In-site

20. As fast as you may dream of…
The above figure (from Microsoft© Azure © control panel) represents a typical day of usage of Fetch News© semantic crawling
processing.
 As shown in the figure…
 Exactly 129.64 Mega Bytes of sources’ contents were captured, analyzed, and converted to a structured text of 41.02 Mega
Bytes in 281.96 seconds –i.e. about 500 Kilo Byte per second including connectivity time..
 Maximum memory usage was 129.75 Mega Bytes fulfilling the needs of 64 parallel threads running on a single core
processor.

21. Proudly introducing
Querying the text streams

22. Significs© Main focus…
CONVERTING
UNSTRUCTURED DATA
TO STRUCTURED DATA
AND
EXTRACTING MEANINGFUL INDICATORS
FOR THEIR CONTENTS WITH
DISTINCT PERFORMANCE
w.r.t various areas of significance.

23. Analysis Dimensions…
Descript
or
Modifi
er
Indecenci
es
Prover
bs
Danger
Safety
Liking
Disliking

24. Main Dimensions Foundations…
Liking
Disliking
Safety
Danger
Basic
Dimensions
Time
Place
Subject
Object
Type
Component
Feature
Descriptors
Modifiers
Indecencies
Proverbs
Sentiment
Dimensions
Needs
Feelings
Actions
Attitude
Object
Dimensions
Descriptor
Dimensions

25. Deep Analysis vs. shallow…
 Shallow Analysis does not depend on the specific
linguistic rules of the target language; thus can’t
guarantee the precise understanding of the textual
contents.
 Deep Analysis uses specific linguistic rules of the target
language; accordingly maximizes the accuracy of the
semantic analysis results.
Info Arab® utilizes its wealth of linguistic technologies (e.g.
speller, morphing, grammar, thesaurus …etc.) in order to
avail a robust deep analysis engine.

26. 3 Levels of Taxonomies…
Generic
Taxonomy
Basic Objects’ types
Basic Features
Basic Components
Basic Measures
Basic Actions
Basic Feeling
Basic Needs
Basic attitudes
…etc.
Domain
Specific
Domain
specific
• Object
• Features
• Measures
• Actions
• Feelings
• Attitudes
D• o…meatcin Specific…
•Subjects
•Issues
•…etc
User
Specific
Taxonomy
User specific
• Object
• Features
• Measures
• Actions
• Feelings
• Attitudes
• …etc
User Specific…
•Subjects
•Issues
•..etc

27. Supported languages…
Significs© is currently able to analyze the following textual language…
• Standard formal Arabic.
• Egyptian common slang.
• Transliterated Arabic texts (written in Latin characters).
• Standard formal English.
• GCC slangs.
Significs© implements the Deep analysis approach to handle languages marked in
BLUE while Shallow Analysis to handle languages marked in GREEN. Future
versions of Significs© are planned to extend the deep analysis algorithms to the other
languages marked in GREEN.

28. up to 20x1012 Areas of Significance
Areas of significance represent the exact requirements of a certain user/customer. They
could be represented in one or more (Correlated/Intersected) dimension.
For example…
 A celebrity may be interested to have a look on any mention about his/her objects
(Persons, Places …etc.).
 A company may want to automatically route all incoming mail that Needs Support of a
certain Product to one of its affiliate companies and measure the customers Satisfaction.
 A marketing agency may be interested to check the Feelings and Attitudes about a certain
Component or Characteristic of one or all of its customers’ products.
 A security agency may want to extract one or all Actions within a certain Subject that
exceeds a certain threshold.

29. Available on almost all platforms…
Significs© is available on the following formats…
• Microsoft© Windows© .DLL.
• MAC OSX, UNIX, LUNIX .LIB & .SO.
• Web service.
• WCF web service.
• http request (with some limitations regarding input text size).

30. Unprecedented Performance…
• 500’000 to 4’000’000 word per second depending
on the analysis depth. Speed
• Total data dictionaries sizes ranges from 20-30
Mega Bytes. Storage
• Core memory is about 80 Mega Bytes.
• Typical memory per thread is about 2 Mega Bytes. Memory

31. Economic feasibility at a glance…
$0.35
$0.30
$0.25
$0.20
$0.15
$0.10
$0.05
$-
For both Fetch-Cloud© and Significs© site licensing
COST PER FEED ANALYSIS
Manual cost Standard Professional Ultimate Open
Cost per feed
Number of feeds
$0.001 / feed @ 200
Million feeds.
Edition
Breakeven
Number of
feeds
Standard 150K
Professional 300K
Ultimate 700K
Open license 3.5M

32. Thank you
Please feel free to contact us for
any enquiry.
We shall be delighted to arrange
for a life demo at any time of
your choice.
Catering for the Arabic language needs in the Digital
era
Info Arab®, Arabic Information Systems, Alaa S. Al-Agamawi & Co.
7 Mosadek Street, Dokki,, Cairo 12312, Egypt.
Phone: +2 02 37 61 38 33 - Fax: +2 02 37 61 38 39 x115
email: info@info-arab.com
www.info-arab.com www.info-arab-technologies.com