PayItForwardThe phrase “pay it forward” describes a situation where the beneficiary of a good deed is asked only to perform a similar good deed for the benefit of someone else. Although it is a centuries-old concept (see Wikipedia for quotes from Benjamin Franklin, Ray Bradbury, Robert Heinlein and others), it was popularized again in the 2000 movie of that name starring Kevin Spacey, Helen Hunt, and Haley Joel Osment.

Pay It Forward - Personal v01A key part of the concept of pay it forward is that individual good acts can have a ripple effect throughout the world.

Rolling intelligence is the enterprise-level equivalent of pay it forward. Effort invested in one business unit, functional area, or type of document storage to visually classify documents benefits the other units, areas, or types of storage that are processed later on. It is the gift that keeps on giving. The reason is that there is a heavy overlap on the types of documents that are used or stored in different areas of an enterprise.

Pay It Forward - Enterprise v01For example, classifying all of the documents that are located on an enterprise’s file shares will also operate to classify many of the files that are stored in SharePoint and that appear as attachments in email stores. Classifying scanned paper archives operates to automatically classify the electronic files that were used to print those documents in the first place.

To a certain extent, the ability to roll the intelligence forward from initial document classification initiatives takes some of the pressure off making the decision about where to start getting a handle on document or content management in an organization – no matter where you start, the intelligence can be rolled forward to subsequent areas.

Pay It Forward - Enterprise v02The intelligence that can be rolled forward includes not only the classification decisions (i.e., what clusters of visually-similar documents to keep as records, and how to label them with document-type names), but also the zonal attribution rules used to identify and extract key attributes or data elements from within documents in specific clusters.

Similarly, the intelligence about how to determine logical document boundaries within compound PDFs or within scanned documents with missing or improper logical document boundaries can also be rolled forward to subsequent initiatives.

As explored in an earlier blog posting, rolling intelligence forward can be very useful in assimilating the operations of other businesses that are brought into an enterprise as a result of merger and acquisition activity.

Being able to roll intelligence forward from initial visual classification and zonal attribution initiatives demonstrates two related concepts: persistence and convergence.

  • Persistence indicates that once classification or attribution decisions have been made pertaining to a specific cluster of visually-similar documents, those decisions persist and are applied to all documents that are ever added to the cluster.
  • Convergence describes the phenomenon where, as documents are being clustered or classified, more and more of them fall into clusters that have already been evaluated. At the point of convergence practically all of the incoming documents have already been evaluated.

BeyondRecognition is the developer of and sole source for visual classification technology for enterprise-scale collections of documents. For information about visual classification,


Comments are closed.