Archive for December, 2011

Use Case: Passenger Name List (PNL) for Secure Flight Program

Wednesday, December 14th, 2011

Case Study Summary:

The Passenger Name List application was developed by ai-one for one of the largest airline ground handling services company in the world.

The PNL Matcher is being used by airlines at the JFK, FRA, and ZHR airports to efficiently and accurately match a PNL (Passenger Name List) with the different suspect lists (no-fly list) supplied by official sources such as the U.S. Department of Homeland Security (DHS) Secure Flight Program.

This application uses the core ai-one™ technology in a limited but very effective way.  The challenge in this area is the need to comply quickly with new U.S. DHS requirements to effectively screen passengers before boarding a flight.

The challenge for such an application is, when a ticketing agent creates a ticket for a passenger from a country that does not use the western alphabetic character set and phonetically spells the name.  Since the spelling could be very different from different agents, the software has to be intelligent enough to find and match suspect passengers to the DHS list.  Additionally phonetic use of characters is varies from country to country but must meet U.S. requirements for quality.

PNL Matcher for Swissport's Secure Flight Program

 PNL Benefits:

  • Fast, very accurate response to a ticket agent
  • Phonetic spelled names can be matched with compliance to all regulations
  • Easy to add new lists
  • Works in any language or character set (language agnostic)

Topic:

Visualize all associations within written text

Kind:

Custom implementation of Topic-Mapper API

Status:

Solutions installed runs since its implementation in 2007 productive
Partner:

Swissport International

Rico Barandun, Product Manager

Application areas:

  • Pattern Matching
  • Security

Target Industries

  • Transportation
  • Homeland security
  • Law enforcement

ai-one Use Case: Enhance OCR of Credit Card Receipts using Machine Learning API

Wednesday, December 14th, 2011

OCR Correction using ai-one machine learning API


Use Case Summary:

The BON Matcher is an ai-one implementation enabling a leading swiss retail store to analyze all scanned credit card receipts.

After the scan process, all credit card receipts are analyzed and matched against patterns using a-one’s API.

Our solution corrects the errors of the optical character recognition (OCR) system when it fails to recognize 100% of the elements.

This was an early validation of our technology. It  affirmed ai-one’s superiority over alternative artificial intelligence-based solutions as a much faster, better quality, and less expensive solution. The retail chain saved substantial operating costs by automating this process and was able to reduce its workforce by 15 people.

The project was finished after 3 months of development time and is still being used for more than 80 stores.

The feature of the technology used in this application is commonly used in document archiving systems where users need to search for documents that have been scanned with many character errors.

Benefits:

  • Improved OCR performance from 80% to 98% in less than a week after implementation.
  • Enhancing OCR recognition in a separate, low-cost post processing process
  • Faster data availability
  • Additional fraud detection possibilities

Deployment:

Customize software development

Status:

Solution in place. Successful since 2006 launch.

Partner:

Swiss Data Safe AG

Application areas:

  • OCR recognition
  • Numerical series matching
  • Data management / Archiving

Target Industries:

  • Information management
  • Retail

 

OCR Correction Workflow Using Machine Learning API