Business Process Optimization for Transaction Processing

Business Process Optimization for Transaction Processing

Transaction Processing is a horizontal capability of Xerox Services which deals with business process form digitization. The form digitization process comprises of the Scan, OCR and the Data Entry (DE) tasks which are controlled by an automated workflow. This project aims to optimize this form digitization process through the works done in the following projects:

Crowdsourcing for Transaction Processing

The Transaction Processing capability currently performs the DE process by outsourcing scanned form images and OCR data to offshore Xerox Services locations for workers to verify or edit the OCR data or key the form data correctly. Outsourcing is losing its cost advantage owing to resource set up and management cost (IT and Human), attrition and inability to scale up / down the workforce as per the workload etc.

This project uses crowdsourcing as an alternative to outsourcing to overcome the above limitations while optimizing digitization cost. This solution handles challenges relating to crowdsourcing business process tasks viz. confidentiality and compliance issues in using the crowd for DE, ensuring quality of digitization while using untrained crowd to perform the DE task and completing the task throughput on time while using an uncommitted workforce.

Form Correction for Improved OCR

Business Process forms processed by the Transaction Processing capability are filled and submitted by different people and organizations and contents falling outside field boundaries are common. The OCR read of the forms are limited to a great extent based on the positioning of contents within the field zones of the form. In addition, the scanning noise renders variation in the size of the form image which causes the field zone calibrations to change across forms – while the OCR works on a standard field zone calibration. This is in turn limits the OCR read quality. The presence of stamps in forms adds to the OCR limitation.

This project uses a knowledge based approach to identify contents in the form and associate them to their correct field zones. It then lifts and shifts the contents to a new form positioning the contents appropriately within field zone boundaries. It also identifies and eliminates stamps from being moved to the new form. This new form when subject to OCR yields a better read, thereby helping to helps to minimize the DE effort that follows OCR. The minimization of human effort helps to optimize the cost of digitization.

Patents

  1. 20110991-US-NP - Hybrid Multi-Iterative Crowdsourcing System
  2. 20120199-US-NP- Method And System For A Text Data Entry From An Electronic Document

Patents Filed

  1. 20120142-US-NP -Method And System For Processing Application Forms
  2. 20111593-US-NP - Method And System For Recommending Crowdsourcability Of A Business Process
  3. 20120378 US-NP – Crowdsourcing Directory System
  4. 20121434US01 Methods and System for Crowdsourcing a Task
  5. 20121274US01 Methods and Systems for Creating Tasks of Digitizing Electronic Document