OCR (Optical Character Recognition)

Document Automation Software Solution, Mortgage Industry Needs

The mortgage industry heavily relies on paper-based processes. Mortgage documents are generated at various stages in the mortgage value chain and come from multiple vendors and partners. There are custom forms as well. The industry has embraced imaging and is now ready to automate document-to-data entry and integrity aspects of the business. Many mortgage companies have chosen to offshore the data entry and document-to-data integrity processes. Visionet has been providing similar offshore data entry services to the mortgage industry for a number of years and has compiled comprehensive data on human versus OCR metrics specific to the mortgage vertical.

Visionet studied three such mortgage processes in detail and has found that OCR technology produced mixed results. The areas of Visionet’s focus are:

1. Legal Descriptions for Lien Releases: Over six million existing properties change hand every year. Approximately twenty percent of these releases require a legal description to be entered into lien release software such as VisiRelease® . Visionet has compiled a comprehensive dictionary of the most used words in legal descriptions and has witnessed a 30% boost in the number of Releases/Legal Descriptions completed per person.

2. 1003 Data Entry: Every mortgage application has a 1003 form. Currently some customers require 100% blind data entry and then compare the entered data against the data from the origination system. Some customers use statistical sampling to minimize costs. The data on 1003 is critical for reporting purposes. The 1003 is a complex document containing text, check boxes and data. Visionet has found OCR to be less effective in terms of the net productivity gain when compared with 1003 manual data entry. The accuracy and productivity improved when Visionet employed two commercially available OCR engines in tandem to highlight the mismatches. This technique demanded heavy computing and image manipulation. Most businesses have ample computing power, as desktop are idle during the night. The mortgage companies should explore this method to improve doc-to-data integrity before rushing offshore.

3. Mortgage Document Indexing: When mortgage companies image various mortgage documents at various stages of application processing they often place a header page to index individual documents. The alternative, which Visionet’s clients have used over the years, is to scan the whole file into a single bucket and then use offshore resources to visually inspect each document and electronically index them. Visionet has found OCR technology to provide decent productivity boost.

Visionet uses a combination of data extraction proprietary OCR tools and dictionaries to populate the necessary data. This reduces cycle times and significantly lowers cost.

Document Indexing through OCR
Reduce data entry costs and improve accuracy of data of documents e.g.

• Legal Description
• Loan Application – 1003
• Note
• HUD-1
• Title Policy
• …

VisiOCR^®- Approach

• OCR Toolkit independent design
• Industry leading OCR Toolkit integration
• VisiLoanReview configurable document templates
• Exception processing based on ‘read’ scores
• Minimizes data entry requirements
• Automates document validation

OCR Tools

VisiOCR^® uses industry standard OCR toolkit and Visionet has developed a number of ‘tools’ to improve recognition

Anchors

Image or Text identifiers used to recognize documents or sections within documents
Heuristics based text anchors recognition
Dictionary based target text recognition

Auto Detection of Documents

OCR assisted page/document recognition and indexing
Template coordinates alignment and compensates for scanning errors – skew, speckle etc.
Supports multiple image resolutions

VisiTrak^® - Technical Architecture

Utilizes Existing Imaging and Scanning infrastructure
Integrates with document imaging systems

Batch OCR Process

Reduces interactive processing
Lower licensing fees for OCR engine
Exceptions based corrective actions

API Based Integration

Source documents based OCR Parameters
XML based or tables based OCR output

Pre-defined Document Templates

Auto recognition of documents/pages
Flexible templates allow adjustment according to image quality
Mortgage “Dictionary” to improve recognition

Integrated with VisiLoanReview^® , VisiRelease^® and other Visionet Workflow products