# US 12,008,621 B1 · Search query processing system > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,008,621 B1 - Document ID: US12008621B1 - Status: Granted - Kind: B1 - Application number: 18/328,681 - Filed: 2023-06-02 - Granted: 2024-06-11 - Earliest claimed: 2023-06-02 - Assignees: InstaProtek Inc. - Applicants: InstaProtek Inc. - Inventors: Farsheed Atef - Web page: https://patentgrep.com/patent/US12008621B1 ## CPC - G06Q 30/0625: {by formulating product or service queries, e.g. using keywords or predefined options} - G06F 16/953: Querying, e.g. by the use of web search engines - G06Q 30/0282: Rating or review of business operators or products - G06Q 30/0639: {Locating goods or services, e.g. based on physical position of the goods or services within a shopping facility} - G06F 40/30: Semantic analysis - G06Q: INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract Systems and methods for enhanced search query systems and methods are described. A large language model is used to extract metadata from a plurality of items of content related to respective items. Sentiment analysis is utilized to determine user sentiment regarding respective item attributes and/or items. Values are associated with a given item's attributes using the determined user sentiment. A user query is received over a network. A search is performed for items satisfying the query using the extracted metadata. Items identified in the search may be ranked using respective values, the values based at least in part on the determined user sentiment, assigned to respective attributes associated with the items. The ranked search results may be transmitted to the user device, the ranked search results corresponding at least in part to the item and/or item attribute rankings, wherein the ranked search results are displayed via the user device. ## Claims ### Claim 1 A search system configured to perform searches in response to user queries, the system comprising: a network interface; at least one processing device operable to perform operations comprising: access a plurality of items of content comprising text related to one or more items; access helpfulness data with respect to respective items plurality of items of content and based at least in part on the accessed helpfulness data determine whether at least a portion of the items plurality of items of content are to be excluded from sentiment analysis; tokenize the text of at least a portion of the plurality of items of content related to one or more items, wherein certain portions of the text are split into individual words as respective tokens, and at least one multi-word expression is maintained as a single token thereby preserving its semantic meaning; utilize a large language model to extract metadata from at least a portion of the tokenized plurality of items of content related to one or more items, wherein a performance of the large language model is improved by training the large language model on a specific task of sentiment analysis using labeled data; perform sentiment analysis to determine user sentiment regarding respective item attributes and/or items; assign respective values to one or more of a given item's attributes based at least in part on the determined user sentiment; receive a handshaking message from a client on a first user device that identifies a cipher suite supported by the client and that includes a string configured to be used in generating an encryption key; generate a key configured to be used in encrypting messages between the system and the first user device; respond to the client with a handshaking signal which identifies the cipher suite suit and encryption version to be used; receive, via the network interface over a network from a first user device, associated with a first user, a first query; enhance a search functionality by: perform a search for items satisfying the first query using the extracted metadata; rank items identified in the search based at least in part on respective values, the values based at least in part on the determined user sentiment, assigned to respective attributes associated with the items; and transmit over the network, using the network interface, ranked search results, to the user device, the ranked search results corresponding at least in part to the item and/or item attribute rankings, wherein the ranked search results are displayed via the user device. ### Claim 2 (depends on 1) The system as defined in claim 1, the operations further comprising: aggregating data related to specific metadata; determining sentiment information using the aggregated data; generating rating data for item attributes corresponding to the specific metadata using the sentiment information. ### Claim 3 (depends on 1) The system as defined in claim 1, wherein the large language model comprises one or more hidden layers and one or more pooling layers. ### Claim 4 (depends on 1) The system as defined in claim 1, wherein the one or more items comprise respective products. ### Claim 5 (depends on 1) The system as defined in claim 1, the operations further comprising identifying feedback for a given item of content, and based at least in part on the identified feedback, determine whether metadata and/or sentiment analysis from the given item of content is to be used in ranking items identified in the search. ### Claim 6 (depends on 1) The system as defined in claim 1, wherein the plurality of items of content related to one or more items comprises user item reviews and/or an aggregate rating based on the item reviews. ### Claim 7 (depends on 1) The system as defined in claim 1, wherein prior to utilizing the large language model to extract metadata from the plurality of items of content related to one or more items, a data cleaning process is performed on the plurality of items of content to remove HTML tags, punctuation, and/or stopwords. ### Claim 8 (depends on 1) The system as defined in claim 1, wherein the ranked search results comprise images of the corresponding items, selected excerpts from corresponding user content, and/or sentiment values. ### Claim 9 (depends on 1) The system as defined in claim 1, wherein the extracted metadata comprises respective identifiers associated with respective items, respective content creation dates, and respective item features. ### Claim 10 (depends on 1) The system as defined in claim 1, the operations further comprising: scraping at least a portion of the items of content from a plurality of websites and/or databases. ### Claim 11 A search system configured to perform searches in response to user queries, the system comprising: a network interface; at least one processing device operable to perform operations comprising: access a plurality of items of content comprising text related to one or more items; access helpfulness data with respect to respective items plurality of items of content and based at least in part on the accessed helpfulness data determine whether at least a portion of the items plurality of items of content are to be excluded from sentiment analysis; tokenize the text of at least a portion of the plurality of items of content related to one or more items, wherein certain portions of the text are split into individual words as respective tokens, and at least one multi-word expression is maintained as a single token thereby preserving its semantic meaning; utilize a large language model to extract metadata from at least a portion of the tokenized plurality of items of content related to one or more items, wherein a performance of the large language model is improved by training the large language model on a specific task of sentiment analysis using labeled data; perform sentiment analysis to determine user sentiment regarding respective item attributes and/or items; assign respective values to one or more of a given item's attributes based at least in part on the determined user sentiment; receive a handshaking message from a client on a first user device that identifies a cipher suite supported by the client and that includes a string configured to be used in generating an encryption key; generate a key configured to be used in encrypting messages between the system and the first user device; respond to the client with a handshaking signal which identifies the cipher suite suit and encryption version to be used; receive, via the network interface over a network from a first user device, associated with a first user, a first query; enhance a search functionality by: perform a search for items satisfying the first query using the extracted metadata; rank items identified in the search based at least in part on respective values, the values based at least in part on the determined user sentiment, assigned to respective attributes associated with the items; and transmit over the network, using the network interface, ranked search results, to the user device, the ranked search results corresponding at least in part to the item and/or item attribute rankings, wherein the ranked search results are displayed via the user device; receive a selection of a first item in the search results; determine a location of the first item; determine a first location of the first user; cause a first navigation route to be generated from the first location of the first user to the location of the first item; and cause the first navigation route from the first location of the first user to the location of the first item to be transmitted to and displayed by the first user device in association with a map. ### Claim 12 (depends on 11) The system as defined in claim 11, the operations further comprising: aggregating data related to specific metadata; determining sentiment information using the aggregated data; generating rating data for item attributes corresponding to the specific metadata using the sentiment information. ### Claim 13 (depends on 11) The system as defined in claim 11, wherein the large language model comprises one or more hidden layers and one or more pooling layers. ### Claim 14 (depends on 11) The system as defined in claim 11, wherein the one or more items comprise respective products. ### Claim 15 (depends on 11) The system as defined in claim 11, the operations further comprising identifying feedback for a given item of content, and based at least in part on the identified feedback, determine whether metadata and/or sentiment analysis from the given item of content is to be used in ranking items identified in the search. ### Claim 16 (depends on 11) The system as defined in claim 11, wherein the plurality of items of content related to one or more items comprises user item reviews and/or an aggregate rating based on the item reviews. ### Claim 17 (depends on 11) The system as defined in claim 11, wherein prior to utilizing the large language model to extract metadata from the plurality of items of content related to one or more items, a data cleaning process is performed on the plurality of items of content to remove HTML tags, punctuation, and/or stopwords. ### Claim 18 (depends on 11) The system as defined in claim 11, wherein the ranked search results comprise images of the corresponding items, selected excerpts from corresponding user content, and/or sentiment values. ### Claim 19 (depends on 11) The system as defined in claim 11, wherein the extracted metadata comprises respective identifiers associated with respective items, respective content creation dates, and respective item features. ### Claim 20 (depends on 11) The system as defined in claim 11, the operations further comprising: scraping at least a portion of the items of content from a plurality of websites and/or databases.