FIRST LOOKS:

'Thumbprint' Automation

By Doug Henschen

When it comes to big-league conversion projects, most of the cost is usually tied up in document prep and indexing labor. Amitech (www.amitech.com) has developed the TurboScan 2000 capture system to automate records such as patient, insurance, mortgage or personnel files with as little batch preparation as possible.

Whether you’re dealing with paper records or microfilm, TurboScan lets you set up document “thumbprints” that spot document types without the barcodes or separator sheets commonly employed by other systems. Not to be confused with form ID, this pattern-recognition approach can be applied to unstructured correspondence as well as fixed and variable forms.

Rather than relying on rigid anchor points that may not exist on your legacy documents, TurboScan gauges the overall look and feel of a particular document type from a thumbnail-size image. If the application permits, you can also rely on more conventional barcodes, patch codes, blank pages or image file names. The system supports seven levels of batch hierarchy, so you can specify complex structures such as state, city, office, room, file cabinet, folder, document and page.

Like most high-volume capture solutions, TurboScan is a three-tier, 32-bit system that places a job/batch manager on a NT server and up to nine processing modules on one or more Windows 9X workstations. You can choose from an Access, SQL Server or ODBC database, and everything’s networked together with IP protocols, so you can handle processes remotely. Service bureaus, for example, can perform scanning at the client site if documents can’t leave the premises.

At a Glance Product: TurboScan 2000 Vendor: Amitech, Springfield, VA, 703-256-2020 Description: Paper and microfilm capture system for high-volume conversion projects, with 32-bit, client/server architecture and nine processing modules that enable load balancing and distributed processing with IP networking. Strengths: Pattern recognition identifies a broad range of document types without special anchor points or recognition zones. Specialized microfilm image processing and index conversion. Supports seven levels of batch hierarchy. Administrative stats support service bureaus. Weaknesses: Complicated installation and setup procedures require initial vendor support. Database lookups and other validations require coding. Lacks integrated PDF conversion. Client support: Windows 9X/NT Server support: Windows NT 4.0 Database support: Access, SQL, ODBC Price: $3,995 per client ProductInfo 201

The processing modules cover the usual bases — Capture (scan/import), Enhance-ment, Automatic QA (quality assurance), Separation, ManualIndex, Full-text OCR, Verification and Export. AutoIndex is where the “thumbprint” recognition method comes into play. Once an image is matched to a document type, the system can start filling index fields from OCR or ICR zones set up on the template. You can code database lookups to validate results, and the system can even go backward and forward to endorse every image within a batch confirmed to relate to a particular employee, client or other grouping.

In our preliminary tests, TurboScan’s thumbprint ID feature worked best with documents that were consistent within type and dissimilar from one another. For example, two different types of questionnaires that shared similar header and footer paragraphs weren’t always distinguished from each other. When one of these questionnaire types was mixed with very different-looking magazine feature manuscripts, the results were much better. Press releases generally contain consistent types of content, but presentation and format vary widely; we found it difficult to consistently identify them in a mixed batch with article manuscripts.

If you’re converting from microfilm, Amitech has specialized image processing that will cope with the messy borders and degraded images often found on film. TurboScan can also read the one- to three-level “blip” indexes that often appear along the edge of roll microfilm.

Amitech is targeting TurboScan at service bureaus and other organizations facing 500,000+page conversion projects up to 1 million+per-day processing volumes. The software itself is very competitive at $3,995 per client. This includes all nine processing modules, TWAIN and ISIS support and an unlimited-use license. The company sells directly, and it generally spends a week or two helping customers to install software, access automation and data capture opportunities, and define and tune document templates.

If you’re dealing with big volumes and particularly if you’re converting microfilm, TurboScan 2000 should be on your short list. If its unique pattern recognition approach can yield even incremental gains in automation, you’ll stand to gain big savings of time and money.