Recognition Quality Management

A system for improving recognition quality by reducing a total number of a plurality of objects in a database, comprising:

• a. a plurality of data structures each pertaining to one of the plurality of objects;
• b. a recognition quality routine, wherein the recognition quality routine comprises:
• i. a string quality score (SGS) subroutine configured to concatenate a plurality of fields containing strings in each of a plurality of data structures pertaining to the object, and calculating a SGS based on the similarity of the concatenated strings;
• ii. an input quality score (IQS) subroutine configured to calculate an IQS for the object based on a weighted measure of a plurality of input overall quality scores and input confidence scores;
• and
• iii. a link quality score (LQS) subroutine configured to:
• 1. calculate an LQS for the object using comprising the number of unique links, a number of data structures associated with the object, a count of the unique link that occurs most often, and a ratio of the number of unique links to the number of data structures;
• and
• 2. determine if any of a plurality of LQS exceptions apply and thereby override the previously calculated LQS for the object by replacing the previously calculated LQS for the object with a pre-defined LQS value associated with such LQS exception;
• iv. a group quality grade (GQG) subroutine configured to:
• 1. calculate a GQG for the object based on a weighted measure of the SGS, IQS, and LQS, wherein the GQG is a single numerical value falling within a defined range;
• and
• 2. applying a plurality of GQG exceptions to the GQG score for the object to determine if any of the GQG exceptions apply and override the previously calculated GQG for the object by replacing the previously calculated GQG for the object with a pre-defined GQG value associated with such GQG exception to produce a final GQG (FGQG);
• and
• v. a cut-off threshold routine configured to apply a cut-off threshold FGQG value to the database such that objects with an FGQG below the cut-off threshold FGQG value are suppressed in the database, wherein the total number of objects in the database after application of the cut-off threshold routine is less than the total number of objects in the database prior to application of the cut-off threshold routine.

A method for improving recognition quality, comprising the steps of:

• a. concatenating a plurality of fields containing strings in each of a plurality of data structures in a database each pertaining to one of a plurality of objects, and calculating a string quality score (SGS) based on the similarity of the concatenated strings;
• b. calculating an input quality score (IQS) for the object based on a weighted measure of a plurality of input overall quality scores and input confidence scores;
• c. calculating a link quality score (LQS) for the object, with inputs to the LQS calculating comprising the number of unique links, a number of data structures associated with the object, a count of the unique link that occurs most often, and a ratio of the number of unique links to the number of data structures;
• d. calculating a group quality grade (GQG) for the object based on a weighted measure of the SGS, IQS, and LQS, wherein the GQG is a single numerical value falling within a defined range;
• and
• e. repeating the foregoing steps for a plurality of objects in the database, and applying a cut-off threshold FGQG value to the database such that objects below the cut-off threshold FGQG value are suppressed in the database,
• wherein the plurality of objects in the database after the step of the application of the cut-off threshold FGQG value comprises at least three percent (3%) fewer objects than prior to application of the method steps hereof.