Technology groups documents based on an image, not textual, analysis, permitting rapid development of cluster-specific, location-based data extraction rules
GERMANTOWN, TN, April 24, 2012 /24-7PressRelease/ -- John Martin, founder and CEO of BeyondRecognition, LLC, announced today that BeyondRecognition ("BR") has developed software to cluster like documents based on an analysis of document images and to then quickly build cluster-specific, location-based data extraction rules. Those rules combined with BR's innovative glyph-clustering text recognition process permit the extraction of fields of data about a document's attributes or contents
Martin predicted, "Image-based document clustering and the graphical interface for the rapid development of cluster-specific data extraction rules provides a level of accuracy and speed that rivals and in some cases exceeds manual coding. This type of approach may dramatically reduce the volume of basic, objective manual document coding work being done, especially with systems like ours that programmatically detect document boundaries."
Martin also noted, "We're seeing a great deal of interest in this approach in the mortgage and energy sectors. The mortgage industry in particular typically has a relatively finite number of documents in loan files supporting the loan decisions, with definable types of data being of interest on each type of document. Our process could greatly lower the cost of tracking all those data elements during loan initiation, or to quality control the file for audit or sale purposes."
Focus Data Management President and Founder Michael Mulcahy states " text analysis or retrieval system can’t find what isn’t in its index- rendering the knowledge contained therin less than accessible. "
Barbara Johnson, CFA, former executive of USAA Federal Savings Bank, serving as Chief Credit Officer and Senior Vice President of Real Estate Lending Services and now a Principal with Saccadent, a Financial Consulting firm, has reviewed the clustering and data extraction capabilities and offered the comment that, "In today's environment the ability to extract, utilize and match data across a variety of documents is incredibly powerful. This type of technology offers the promise of significantly decreasing the time and cost to process a thoroughly compliant loan from application through origination, audit, sale and servicing. An automated system to confirm all the critical items match throughout the process and are in the appropriate format and location on all documents would be invaluable. Lending is a document-rich industry and the time is perfect for this type of technology."
In the energy sector, BR's glyph clustering technology would also make it possible to search for symbols used on maps to indicate things like radioactive wells, salt-water wells or API number codes.
Martin commented about the BeyondRecognition product development road map, "Because we track the page position of each character and word we will at some point develop true proximity searching where users could specify how many inches or centimeters either vertically or horizontally the search terms needed to be in relation to each other in order to satisfy the search parameters. This could include a relational operator for looking in the same row or the same column or within 'x' rows or 'y' columns. Our standardized output files would also enable search engine providers to incorporate this functionality in their engines."
Regarding electronic discovery, Martin said that, "Some of the e-discovery experts who have reviewed the system have suggested that it might be worthwhile to convert native e-discovery files to image format just to be able to use BR's image-clustering and location-based data extraction capabilities. The BR true proximity/relational search capabilities would provide another reason for taking electronic files to image."
Persons interested in receiving updates about BeyondRecognition should send an email to Info@beyondrecogntion.net.
About BeyondRecognition
BeyondRecognition is a "textnology" company that has developed unique character, word and document attribute recognition and extraction capabilities for analyzing image-based documents. Disclosure of further details is being deferred until one or more patents on the process are filed. BeyondRecognition is working with a select number of companies in the electronic discovery and document management industries. Due to bandwidth considerations, BeyondRecognition will be unable to give detailed consideration to further requests or inquiries until mid-June, 2012. If your company would like to explore licensing or partnership opportunities with BeyondRecognition in that time frame, please email John Martin at jmartin@beyondrecognition.net providing your contact information and a summary of what your interests are.
No comments:
Post a Comment