EnPRO Business Plan Presentation

1
EnPRO Business Plan Presentation

• Arabic Entity Tagger

2
The Team

• Faculty advisor Dr. David Grossman
• Project Leader Michael Lee
• Assistant Project Leader Syed Uzair Aqeel
• Team members
• Steven Wadhwani (CS, Senior)
• Naveed Aziz (CPE, Senior)
• Tufail Baloch (EE, Sophomore)

3
The Problem

• Existing search does not do any entity tagging.
  Queries simply give results but do not label the
  important parts of a results.
• Names
• Addresses
• Phone
• Also, large volumes of text need to be mined for
  interesting information.

Database Of Names Organizations Etc.
Text Mining
Pile of Documents
4
Google Today
5
Google Entities
6
Arabic Google
7
Arabic Google entities
8
How We Solve The Problem

• Named entity recognition (NER) is an important
  step towards semantic understanding of text
• We developed and tested a method for recognizing
  and extracting named entities in Arabic

9
The Product

• Our tagger is capable of recognizing named
  entities in Arabic text
• Person, e.g. President Bush
• Location, e.g. White House
• Organization, e.g. ACM
• Miscellaneous, e.g. December 25, 2003
• A combination of Arabic-specific features and
  state-of-the-art data mining models are used for
  entity recognition

10
Customers
11
Market Size

• Growing Demand for Software and content in
  Arabic.
• Companies that are poised to take advantage of it
• 42 multi-nationals developing software in Arabic
• Search engine developers
• Banks
• publishers.
• 17 major business organizations operating in
  Arabic Countries.

• Government agencies throughout the world
• FBI
• Department of Homeland Security
• United Kingdom Ministry of Defense.
• Law enforcement Agencies in the Middle East

12
Product Strategy

• Distribute as a software library
• Search Engine Vendors can call this library to
  make Googleentities
• Distribute as a plug-in to applications
• Install on client machines with existing software
  packages that support plug-ins
• For example (Microsoft Outlook entities,
  Eudoraentities)
• Stand alone Application
• Customize based on user preference

Database Of Names Organizations Etc.
Text Mining
Pile of Documents
13
Marketing Strategy

• Niche market focused advertising
• Dispatching product brochures to potential
  customers
• Trade Shows (Ex SearchEngine Strategies)
• Demos
• Online demonstration for potential customers

14
Competitive Landscape

• After fairly strong search, did not find another
  Arabic Entity taggers commercially available.
• Possible future competition
• Numerous English Entity Taggers
• ClearForest
• Attensity
• SRA (Netowl)
• Lockheed Martin (Aerotext)
• CICERO
• TextWise
• Inxight

15
Expenses
On Off expenses for Year 1
16
Financials
17
Financials

18
Progress

• Brand new IPRO
• Completed prototype Arabic language entity tagger
• Draft business plan
• Tested accuracy of the tagger
• (about 70 of names, organizations, locations,
  etc. found)

19
Path Forward

• Continue to improve product
• Effectiveness
• Efficiency
• Might try other Semitic languages

20
Risks

• Depend on Arabic Expertise, need Arabic speaking
  developers
• Arabic world is highly insular
• Relatively low barrier of entry,
• All English taggers could enter the market
• Difficult to protect intellectual property, many
  different approaches are possible.

21
Risk Mitigation

• Obtain significant share of small market
• Ensure that we are high quality
• Keep price lower than the competition
• Multiple Arabic speaking developers.

22
Summary

• We are the first to the market
• Niche market, less inclined to be a major target
  of large companies like Microsoft
• Profitable in three years
• Low investment costs
• Growing Margins
• Big Potential in Government contracts
• New Defense Budget is 401 Billion Dollars
• Source The Economist Nov29th Dec 5th 2003