OCR/ICR Application Process Showing Growing Pains;DMAC Can't Guess Which Approaches & Engines Will Survive"Use of a robust, multi-tasking operating system... is necessary to support image caching plus functionalities such as multiple field display, automated work distribution, and pipelined, hierarchical edit and control procedures."Gingrande develops his theme by listing seven areas which need addressing for OCR/ICR to work. He also summarizes what a hard year of work in "heads up" and OCR/ICR data entry has taught DMAC. Because the industry is nowhere near a complete front to back process, DMAC has backed off from trying to recommend a particular OCR/ICR engine or pretend we have at this time a complete solution. We Accept Where We Are Instead we at DMAC have started working on more standard interfaces for accepting OCR/ICR data, such as accepting comma delineated files in 7.3i, new verbs to properly edit the data, printing support to properly show the data, comfortable alliances with packagers and installers of OCR/ICR engines, and better "heads up" displays, more command line control, etc. Arthur's Summary is Correct We in development subscribe to Arthur Gingrande's list of major factors in making OCR/ICR work for the client. We will be working on those improvements which lie within the Unibase by DMAC environment. Arthur Gingrande's summary of the major factors required to enable cost effective OCR/ICR follows. The complete article in INFORM is worth reading. In parenthesis are Fred's comments about DMAC's state. Intelligent Caching 1. Intelligent caching of field images in RAM. (Arthur thinks ten fields forward and backwards at all times is a good practical limit. Unibase by DMAC currently has in RAM the images associated with two records where users can define a record with over ten fields). Use Multi-tasking OS 2. Use of a robust, multi-tasking operating system. (Arthur says almost everyone agrees a DOS-based, MS Windows environment simply isn't up to the job. We agree, come to TAWPI to see our new platforms run). Multiple User-Defined Edit Zones 3. User-defined, simultaneous display of multiple editing windows. (We agree, look for additions in this area this fall). Use Thresholds 4. Use of thresholds to decide whether or not to key individual OCR rejects versus the whole field. (Thresholds are absolutely needed here and on the level of accuracy of the reject of a character. DMAC is working on this now). Multiple User-defined Edits 5. Multiple, user-defined, hierachical editing structure. ( Unibase by DMAC has new verbs and check boxes coming to expand this area). User-defined Control Procedures 6. User-defined, self-auditing control procedures and automated workload distribution. (DMAC is rapidly adding command line capability for all new features in Unibase by DMAC so DMAC can be controlled from within a user-defined control procedure. We don't dictate what the control procedure is or which vendors product is used; we are supporting several common control systems). Maximum Throughput 7. Maximizing key entry speed is treated as part of a larger throughput issue. (Arthur summarizes all the work need in this area. The Unibase by DMAC speed is right there with other products. ) Arthur concludes his article by stating that the cost-justification entry level must be lowered to create a competitive advantage for the user. (DMAC agrees; we are working on it).
|