Computer readable electronic records automated classification system
View Patent ↗Discloses a system for the automatic classification of electronic documents that are candidates to become an official record. A training data set of documents, each document having a pre-assigned records classification instance, is processed by a classification agent operating in training mode to establish a probabilistic relationship between each classification instance and the contents of a document. The training data set includes at least several documents per classification instance. A trained classification agent operating in evaluation or classification mode receives a candidate document and produces a result. A decision control table contains user specified contiguous result ranges to select one document processing action from a list of document processing actions based on the result produced by the trained classification agent. The list of document processing actions includes further automatic document processing to assign a classification or providing user input capability to enable human intervention and control to classify the candidate document and result in the candidate document being combined with appropriate metadata, and a file plan classification to become an official record.
1. A method of classifying an electronic document in a computer-based system comprising:
generating a classification agent result comprising a plurality of classification instances and for each classification instance a confidence data indicating a degree of confidence that the electronic document is associated with that classification instance; and
selecting based on a user-configurable selection criteria responsive to the classification agent result, the user-configurable selection criteria including at least one criterion based on the confidence data, a processing option from a plurality of processing options including:
assigning without human intervention a classification instance from said plurality of classification instances to the electronic document, wherein the assignment is based at least in part on said confidence data,
producing a list of at least two classification instances for the electronic document and presenting the list to a human user to receive from the human user a selection of a selected classification instance to be assigned to the electronic document, and
assigning a review classification instance to the electronic document to indicate that the electronic document is required to be reviewed by a human user to determine which, if any, classification instance should be assigned to the electronic document;
wherein selecting the processing option for the electronic document further comprises assigning the classification instance without human intervention if at least one classification instance in the plurality of classification instances has associated with it a confidence data that exceeds a first threshold.
2. The method of claim 1 wherein generating the classification agent result comprises:
receiving the electronic document;
receiving at least one classification instance; and
associating the at least one classification instance with the electronic document.
3. The method of claim 1 wherein user-configurable selection criteria includes a numeric range of confidence factors associated with a subset of the plurality of classification instances.
4. The method of claim 1 where producing the list further comprises:
presenting to a user a graphical representation for each of the at least two classification instances; and
selecting one of the at least two classification instances with which to classify the electronic document.
5. The method of claim 4 wherein selecting one of the at least two classification instances further comprises receiving user input:
confirming selection of the one of the at least two classification instances; or
rejecting selection of at least one of the at least two classification instances.
6. The method of claim 1 wherein classifying the electronic document further comprises assigning a null classification instance to the electronic document.
7. The method of claim 1 wherein selecting the processing option for the electronic document further comprises, producing the list and obtaining input from the human user if no classification instance in the plurality of classification instances has associated with it a confidence data that exceeds the first threshold but each of at least a prescribed number of classification instances in the plurality has a respective confidence data that exceeds a second threshold that is lower than the first threshold, and assigning the review classification instance if no classification instance in the plurality of classification instances has associated with it a confidence data that exceeds the first threshold and fewer than the prescribed number of classification instances in the plurality have confidence data that exceed the second threshold.
8. The method of claim 1 further comprising:
determining that no classification instance is relevant to the electronic document; and
assigning a null classification instance to the electronic document.
9. A computer-based system for classifying an electronic document comprising:
a memory; and
a processor coupled to the memory and configured to execute instructions stored in the memory which when executed by the processor provide;
a classification agent configured to provide a classification agent result comprising a plurality of classification instances and for each classification instance a confidence data indicating a degree of confidence that the electronic document is associated with that classification instance; and
a decision control module configured to select, based on a user-configurable selection criteria including at least one criterion based on the confidence data, a module from a plurality of modules comprising:
an automatic-classification module configured to assign without human intervention a classification instance from said plurality of classification instances to the electronic document, wherein the assignment is based at least in part on said confidence data,
a user selection module configured to produce a list of at least two classification instances for the electronic document and present the list to a human user to receive from the human user a selection of a selected classification instance to be assigned to the electronic document, and
a review module configured to assign a review classification instance to an electronic document to indicate that the electronic document is required to be reviewed by a human user to determine which, if any, classification instance should be assigned to the electronic document;
wherein the automatic-classification module is selected to assign the classification instance without human intervention if at least one classification instance in the plurality of classification instances has associated with it a confidence data that exceeds a first threshold.
10. The system of claim 9 further comprising
a user control input device;
at least one electronic document database; and
an electronic document server document operably connected to the at least one electronic document database, which is configured to store the electronic document, and to the user control input device to receive a user control input.
11. The system of claim 9 wherein the confidence data comprises a number representing a probability.
12. The system of claim 11 wherein each one of the one or more user-configurable classification confidence levels comprises a numeric range of potential results of said classification agent.
13. The system of claim 12 wherein a first numeric range of a first user-configurable classification confidence level is contiguous to a second numeric range of a second user-configurable classification confidence level.