Patent Issued for Intelligent routing control (USPTO 11151660): Progressive Casualty Insurance Company

2021 NOV 09 (NewsRx) -- By a News Reporter-Staff News Editor at Insurance Daily News -- A patent by the inventors Esteves, Jeremy (Mayfield Village, OH, US), Harding, Jarrod D. (Mayfield Village, OH, US), Kohli, Kanwar (Mayfield Village, OH, US), Makrucki, Todd (Mayfield Village, OH, US), filed on December 23, 2020, was published online on October 19, 2021, according to news reporting originating from Alexandria, Virginia, by NewsRx correspondents.

Patent number 11151660 is assigned to Progressive Casualty Insurance Company (Mayfield Village, Ohio, United States).

The following quote was obtained by the news editors from the background information supplied by the inventors:

“1. Technical Field

“This disclosure relates to automated agents, and specifically to automated agents that execute document classifications and determine document distributions at the page level and execute updates or modifications in response to the classifications.

“2. Related Art

“The conversion and distribution of physical documents is challenging. Documents come in many forms and contain diverse content. The documents include proof of prior insurance, insurance cancellation documents, credit authorization forms, discount forms, uninsured motorist forms, insurance application packets, etc., and any combination of information. The endless sizes, page orientations, layouts, and formats make it nearly impossible to process and translate documents into standardized forms. Many systems cannot make logical deductions, make logical inferences, or detect incomplete information. The systems do not learn from experiences or analyze contexts.”

In addition to the background information obtained for this patent, NewsRx journalists also obtained the inventors’ summary information for this patent: “A processing system and method (referred to as a processing agent) translate and expedite content delivery of physical and virtual documents. The processing agent’s end-to-end learning system discovers contexts and uses contextual knowledge to efficiently route documents to destinations and perform functions. A detection system accurately detects anomalies, and in some applications, protects backend systems from unsolicited messages asynchronously transmitted to many recipients. An automated job separation system parses documents by breaking and splitting documents into individual pages so that only desired pages are routed to apt destinations. The job separation system enables multiple documents to be automatically processed simultaneously without job separation sheets.

“Some processing agents include export application programming interfaces (APIs) that transmit documents in any desired format to a folder, an address, or a destination that may execute further processing. Some systems transmit commands that enable scripts. Scripts may request information from remote sources, generate and transmit forms to users, notify users of outcomes, render one or more indexing systems and/or automation systems and issue audible or visual alerts to sources or users that signal an error, missing information, or a warning. Some document processing agents execute data compressions that reduce memory and bandwidth use and execute multiple encryptions to maintain security. The document processing agents execute classifications across multiple nodes allowing the processing agents to distribute classification and jobs across local and/or remote computing resources.”

The claims supplied by the inventors are:

“1. A method of classifying documents having insurance data, the method comprising: spinning up a remote serverless application in a system as a result of storing a raw document in a local memory remote from a device; the remote serverless application enables: accessing a plurality of documents in an assessment queue stored in a memory of a device; each of the plurality of documents comprising pages having a predefined destination stored in metadata associated with each document; removing the metadata associated with each document; converting the plurality of documents into a plurality of single physical page documents by splitting each of the plurality of documents into a plurality of files consisting essentially of a single physical page; converting each of the plurality of files into separate recognition vectors that represent information conveyed in each of the single physical pages; receiving an access token asynchronously, the access token being valid for a time period and issued by an off-line access object, the access token being pulled from the off-line access object through an external gateway to authorize a processing agent that enables a workflow: classifying the plurality of files through an additive learning algorithm in which routing data is embedded in second metadata associated with each of the plurality of files while the access token is valid and unexpired; merging the classified plurality of files in response to a plurality of rules based at least in part on the second metadata while the access token is valid and unexpired; routing merged documents to a remote destination independent of the predefined destination stored in the metadata while the access token is valid and unexpired; and processing an insurance parameter automatically in response to the routing of the merged documents through an orchestrator, while the access token is valid and unexpired, that executes a check-pointed event-driven workflow that maintains an application state; and generating a refreshed access token by refreshing the access token when the access token expires or becomes invalid without resoliciting an authorization from the off-line access object; the refreshed access token authorizing the processing agent to complete a function and the workflow; and terminating the processing agent’s execution of the function and the workflow when the refreshed access token expires or becomes invalid; where a plurality of computing resources execute the remote serverless application and the plurality of computing resources are dynamically allocated based on a volume of the raw document.

“2. The method of claim 1 further comprising refreshing the access token when the access token expires or becomes invalid without resoliciting an authorization from a remote object where the plurality of documents comprise email or other digital content.

“3. The method of claim 1 where the method is executed from a machine-readable definition file stored in a non-transitory computer media.

“4. The method of claim 1 further comprising distributing the single physical pages across a plurality of cloud servers.

“5. The method of claim 1 further comprising a computer vision module and a machine learning module that detect and identify the absence of a requirement, including a handwritten cursive signature.

“6. The method of claim 1 further comprising applying a weighting to scalar variables that comprise the recognition vectors based on an active grammar or a predefined document type.

“7. The method of claim 1 where the additive learning algorithm comprises a decision tree.

“8. The method of claim 7 where the decision tree comprises a boosted decision tree.

“9. The method of claim 7 where a time required to train the additive learning algorithm is proportional to a number of splits executed on the decision tree.

“10. The method of claim 7 where the decision tree is trained on a contextual association between words or phrases.

“11. The method of claim 1 further comprising causing the files to undergo a data compression.

“12. The method of claim 1 where the merging of the files is based on at least one Boolean function.

“13. The method of claim 1 where the plurality of rules is stored in individual profiles in a memory which determine when the merging occurs, where the merging occurs, or how the merging occurs.

“14. The method of claim 1 where the plurality of rules is stored in individual profiles in a memory which determines computer file formats or how the merged documents are compressed or mediums distribute the merged documents.

“15. The method of claim 1 further comprising causing an issuance of a warning or an alert in response to the classification.

“16. The method of claim 1 further comprising initiating a script in response to the classification.

“17. A non-transitory machine-readable medium encoded with machine-executable instructions for classifying documents having data, where execution of the machine-executable instructions is for: spinning up a remote serverless application in a system as a result of storing a raw document in a local memory remote from a device; the remote serverless application enables: accessing a plurality of documents in an assessment queue accessed from the local memory and stored in a memory of the device; each of the plurality of documents comprising pages having a predefined destination stored in metadata associated with each document; removing the metadata associated with each document; converting the plurality of documents into a plurality of single physical page documents by splitting each of the plurality of documents into a plurality of files consisting essentially of a single physical page; converting each of the plurality of files into separate recognition vectors that represent information conveyed in each of the single physical pages; receiving an access token asynchronously, the access token being valid for a time period and issued by an off-line access object, the access token being pulled from the off-line access object through an external gateway to authorize a processing agent that enables a workflow: classifying the plurality of files through an additive learning algorithm in which routing data is embedded in second metadata associated with each of the plurality of files while the access token is valid and unexpired; merging the classified plurality of files in response to a plurality of rules based at least in part on the second metadata while the access token is valid and unexpired; routing merged documents to a remote destination independent of the predefined destination stored in the metadata while the access token is valid and unexpired; and processing a parameter automatically in response to the routing of the merged documents through an orchestrator, while the access token is valid and unexpired, that executed a check-pointed event-driven workflow that maintains an application state; generating a refreshed access token by refreshing the access token when the access token expires or becomes invalid without resoliciting an authorization from the off-line access object; the refreshed access token authorizing the processing agent to complete a function and the workflow; and terminating the processing agent’s execution of the function and the workflow when the refreshed access token expires or becomes invalid; where a plurality of computing resources execute the remote serverless application and the plurality of computing resources are dynamically allocated based on a volume of the raw document.

“18. The non-transitory machine-readable medium of claim 17 where the additive learning algorithm comprises a plurality of learning models that are combined to generate a summed output, and interfaces a successive learning model separated from the additive learning model that minimizes a plurality of residuals generated from a preceding learning model.

“19. The non-transitory machine-readable medium of claim 17 further comprising distributing the single physical pages across a plurality of cloud servers.

“20. The non-transitory machine-readable medium of claim 17 further comprising applying a weighting to scalar variables that comprise the recognition vectors based on an active grammar or a predefined document type.

“21. The non-transitory machine-readable medium of claim 17 where the additive learning algorithm comprises a decision tree.

“22. The non-transitory machine-readable medium of claim 21 where the decision tree is a boosted decision tree.

“23. The non-transitory machine-readable medium of claim 22 where a time required to train the additive learning algorithm is proportional to a number of splits executed on the decision tree.

“24. The non-transitory machine-readable medium of claim 23 where the decision tree is trained on contextual associations between words or phrases.

“25. The non-transitory machine-readable medium of claim 17 further comprising causing the plurality of files to undergo a triple encryption.

“26. The non-transitory machine-readable medium of claim 17 where the merging of the is based on at least one Boolean operator.

“27. The non-transitory machine-readable medium of claim 17 where the plurality of rules is stored in individual profiles in a memory, which determine when the merging occurs or where the merging occurs or how the merging occurs.

“28. The non-transitory machine-readable medium of claim 17 where the processing occurs in response to the access token being authorized for a prescribed period of time by the off-line object and the access token being reauthorized.

“29. The non-transitory machine-readable medium of claim 17 further comprising a publishing event associated with the processing the parameter through an internal publish-subscribe system that persists over a systems restart.”

There are additional claims. Please visit full patent to read further.

URL and more information on this patent, see: Esteves, Jeremy. Intelligent routing control. U.S. Patent Number 11151660, filed December 23, 2020, and published online on October 19, 2021. Patent URL: http://patft.uspto.gov/netacgi/nph-Parser?Sect1=PTO1&Sect2=HITOFF&d=PALL&p=1&u=%2Fnetahtml%2FPTO%2Fsrchnum.htm&r=1&f=G&l=50&s1=11151660.PN.&OS=PN/11151660RS=PN/11151660

(Our reports deliver fact-based news of research and discoveries from around the world.)