# US 12,223,441 B2 · Systems and methods for classifying electronic documents > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,223,441 B2 - Document ID: US12223441B2 - Status: Granted - Kind: B2 - Application number: 18/600,109 - Filed: 2024-03-08 - Published: 2024-06-27 - Granted: 2025-02-11 - Earliest claimed: 2013-03-15 - Assignees: TSG Technologies, LLC - Applicants: TSG Technologies, LLC - Inventors: Caimei Lu; Ricky John Prosser; Michael Wayne Bryars - Web page: https://patentgrep.com/patent/US12223441B2 ## CPC - G06F 16/906: Clustering; Classification - G06N 5/04: Inference or reasoning models - G06F 16/81: Indexing, e.g. XML tags; Data structures therefor; Storage structures - G06N 5/025: Extracting rules from data - G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks - G06N 20/00: Machine learning - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract A method of classifying an electronic document from an electronic media source includes generating statistical data from one or more training documents and creating a plurality of classification rules, including creating at least one topic model-based classification rule using the statistical data and creating at least one query-based classification rule using one or more user defined categories and the statistical data. The method further includes classifying the electronic document using the at least one topic model-based classification rule. Example systems for classifying an electronic document from an electronic media source are also disclosed. ## Claims ### Claim 1 A system comprising: a memory having instructions stored thereon; and at least one processor to execute the instructions to: generate statistical data from one or more training documents stored in random access memory (RAM) comprising at least one of blogs, messages, and emails, the one or more training documents comprising positive or negative discussions regarding one of a brand, product, and service and filter the one or more training documents to clean textual content, and create a plurality of classification rules, including creating at least one topic model-based classification rule using the statistical data, the at least one topic model-based classification rule is formatted as an XML file and stored in RAM; evaluate the at least one topic model-based classification rule using a precision equation and a recall equation, the precision equation comprising Precision = N ⁡ ( d c + ) N ⁡ ( d c + ) + N ⁡ ( d c - ) and the recall equation comprising Recall = N ⁡ ( d c + ) N ⁡ ( d c ) wherein N(dc+) is a number of test documents correctly classified to a category C based on the at least one topic model-based classification rule, N(dc−) is a number of test documents incorrectly classified to the category C based on the at least one topic model-based classification rule, and N(dc) denotes a number of test documents that should be classified to the category C; and create at least one query-based classification rule using one or more user defined categories and the statistical data, the at least one query based classification rule is formatted as an XML file stored in RAM. ### Claim 2 (depends on 1) The system of claim 1, wherein the system is capable of connecting to a communication network and one of connected to the communication network and disconnected from the communication network. ### Claim 3 (depends on 1) The system of claim 1, wherein the one or more training documents are at least one of unlabeled, unanalyzed, and unclassified. ### Claim 4 (depends on 1) The system of claim 1, the at least one processor further to translate words in the one or more training documents to a desired language. ### Claim 5 (depends on 1) The system of claim 1, the at least one processor further to stem at least one word in the one or more training documents to a root word. ### Claim 6 (depends on 1) The system of claim 1, the at least one processor further to remove at least one stop word in the one or more training documents. ### Claim 7 (depends on 1) The system of claim 1, the at least one processor further to remove at least one low frequency word in the one or more training documents. ### Claim 8 (depends on 1) The system of claim 1, the at least one processor further to improve an efficiency of at least one of a statistical model and a search engine model by processing the one or more training documents to create the plurality of classification rules. ### Claim 9 (depends on 1) The system of claim 1, the at least one processor further to receive at least one electronic document to be classified. ### Claim 10 (depends on 9) The system of claim 9, wherein the at least one electronic document to be classified is received as streaming data. ### Claim 11 (depends on 9) The system of claim 9, the at least one processor further to process the at least one electronic document to be classified. ### Claim 12 (depends on 11) The system of claim 11, the at least one processor further to send the at least one electronic document to be classified, a topic-term matrix, and the topic model-based classification rules to a topic predictor. ### Claim 13 (depends on 12) The system of claim 12, wherein the topic predictor operates in real-time. ### Claim 14 (depends on 12) The system of claim 12, wherein the topic predictor classifies the at least one electronic document by assigning one or more topical categories to each of the at least one electronic document based on the topic-model based classification rules. ### Claim 15 (depends on 12) The system of claim 12, wherein the topic predictor classifies the at least one electronic document by assigning one or more topical categories to each of the at least one electronic document based on the topic-term matrix. ### Claim 16 (depends on 14) The system of claim 14, the at least one processor further to output the one or more topical categories for each of the at least one electronic document. ### Claim 17 (depends on 15) The system of claim 15, the at least one processor further to output the one or more topical categories for each of the at least one electronic document. ### Claim 18 A method, comprising: generating, by at least one processor, statistical data from one or more training documents stored in random access memory (RAM) comprising at least one of blogs, messages, and emails, the one or more training documents comprising positive or negative discussions regarding one of a brand, product, and service and filtering the one or more training documents to clean textual content, and creating a plurality of classification rules, including creating at least one topic model-based classification rule using the statistical data, the at least one topic model-based classification rule is formatted as an XML file and stored in RAM; evaluating, by the at least one processor, the at least one topic model-based classification rule using a precision equation and a recall equation, the precision equation comprising Precision = N ⁡ ( d c + ) N ⁡ ( d c + ) + N ⁡ ( d c - ) and the recall equation comprising Recall = N ⁡ ( d c + ) N ⁡ ( d c ) , wherein N(dc+) is a number of test documents correctly classified to a category C based on the at least one topic model-based classification rule, N(dc−) is a number of test documents incorrectly classified to the category C based on the at least one topic model-based classification rule, and N(dc) denotes a number of test documents that should be classified to the category C; and creating, by the at least one processor, at least one query-based classification rule using one or more user defined categories and the statistical data, the at least one query based classification rule is formatted as an XML file stored in RAM. ### Claim 19 A non-transitory computer readable medium having instructions stored thereon that, when executed by at least one processor, cause the at least one processor to perform operations, the operations comprising: generating statistical data from one or more training documents stored in random access memory (RAM) comprising at least one of blogs, messages, and emails, the one or more training documents comprising positive or negative discussions regarding one of a brand, product, and service and filtering the one or more training documents to clean textual content, and creating a plurality of classification rules, including creating at least one topic model-based classification rule using the statistical data, the at least one topic model-based classification rule is formatted as an XML file and stored in RAM; evaluating the at least one topic model-based classification rule using a precision equation and a recall equation, the precision equation comprising Precision = N ⁡ ( d c + ) N ⁡ ( d c + ) + N ⁡ ( d c - ) and the recall equation comprising Recall = N ⁡ ( d c + ) N ⁡ ( d c ) , wherein N(dc+) is a number of test documents correctly classified to a category C based on the at least one topic model-based classification rule, N(dc−) is a number of test documents incorrectly classified to the category C based on the at least one topic model-based classification rule, and N(dc) denotes a number of test documents that should be classified to the category C; and creating at least one query-based classification rule using one or more user defined categories and the statistical data, the at least one query based classification rule is formatted as an XML file stored in RAM.