# US 12,010,079 B2 · Message suggestions > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,010,079 B2 - Document ID: US12010079B2 - Status: Granted - Kind: B2 - Application number: 17/570,927 - Filed: 2022-01-07 - Published: 2022-04-28 - Granted: 2024-06-11 - Earliest claimed: 2014-08-27 - Assignees: Google Inc. - Applicants: Google LLC - Inventors: Max Benjamin Braun; Nirmal Jitendra Patel - Web page: https://patentgrep.com/patent/US12010079B2 ## CPC - H04L 51/046: Interoperability with other network applications or services - G06N 20/00: Machine learning - H04L 51/02: using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages - H04L: TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION (arrangements common to telegraphic and telephonic communication H04M) - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract A method may involve, for each of one or more messages that are selected from a plurality of messages from an account: (a) extracting one or more phrases from a respective selected message; (b) determining that a conversation includes the respective selected message and one or more other messages from the plurality of messages; (c) generating a first feature vector based on the conversation, wherein the first feature vector includes one or more first features, wherein the one or more first features include one or more words from the conversation; and (d) generating, by a computing system, one or more training-data sets, wherein each training-data set comprises one of the phrases and the first feature vector. The method may further involve: training, by the computing system, a machine-learning application with at least a portion of the one or more training-data sets that are generated for the one or more selected messages; applying the trained machine-learning application to process an incoming message to the account; and responsive to applying the trained machine-learning application, determining one or more reply messages corresponding to the incoming message, wherein the one or more reply messages include at least one of the extracted one or more phrases. ## Claims ### Claim 1 A method, comprising: applying a trained machine-learning model to process an incoming message to a user account operating within a head-mounted display device, the trained machine-learning model being trained using a training data set generated based on at least one normalized phrase extracted from prior messages sent from the user account and at least one feature vector associated with a set of the prior messages sent from the user account; responsive to applying the trained machine-learning model, determining a reply message by mapping the at least one normalized phrase extracted from the prior messages with the at least one feature vector, the reply message including at least one of the at least one normalized phrase extracted from the prior messages; generating, from the reply message, a relevant reply message to the incoming message; and updating a graphic display of the head-mounted display device with the reply message based on how often reply message is selected or provided for selection. ### Claim 2 (depends on 1) The method of claim 1, comprising ranking the relevant reply message and generating a listing of the relevant reply message for display based on the ranking. ### Claim 3 (depends on 1) The method of claim 1, comprising receiving data input indicating selection of the relevant reply message and, responsively, sending the relevant reply message to another user account in a messaging conversation with the user account. ### Claim 4 (depends on 3) The method of claim 3, wherein receiving the data input includes receiving the data input through a microphone, a touchpad on a wearable computing device, or an eye-tracking sensor. ### Claim 5 (depends on 1) The method of claim 1, wherein the trained machine-learning model being trained includes splitting, by a computing device, the prior messages into the at least one normalized phrase extracted from the prior messages. ### Claim 6 (depends on 5) The method of claim 5, wherein the at least one feature vector includes a first feature vector that includes at least one word associated with the prior messages. ### Claim 7 (depends on 6) The method of claim 6, wherein applying the trained machine-learning model comprises includes the incoming message to generate a second feature vector based on the incoming message, wherein the second feature vector include the at least one word associated with the prior messages. ### Claim 8 (depends on 6) The method of claim 6, wherein the first feature vector comprises a count of the at least one word associated with the prior messages. ### Claim 9 A non-transitory computer-readable medium having stored thereon instructions executable by a processing device, the instructions causing the processing device to perform a method, comprising: applying a trained machine-learning model to process an incoming message to a user account operating within a mobile computing device, the trained machine-learning model being trained using a training data set generated based on at least one normalized phrase extracted from prior messages sent from the user account and at least one feature vector associated with a set of the prior messages sent from the user account, responsive to applying the trained machine-learning model, determining a reply message by mapping the at least one normalized phrase extracted from the prior messages with the at least one feature vector, the reply message including at least one of the at least one normalized phrase extracted from the prior messages, generating, from the reply message, a relevant reply message to the incoming message, and updating a graphic display of the mobile computing device with the reply message based on how often the reply message is selected or provided for selection. ### Claim 10 (depends on 9) The non-transitory computer-readable medium of claim 9, wherein the instructions cause the processing device to perform the method further comprising ranking the relevant reply message and generating a listing of the relevant reply message for display based on the ranking. ### Claim 11 (depends on 9) The non-transitory computer-readable medium of claim 9, wherein the instructions cause the processing device to perform the method further comprising receiving data input indicating selection of the relevant reply message and, responsively, sending the relevant reply message to another user account in a messaging conversation with the user account. ### Claim 12 (depends on 9) The non-transitory computer-readable medium of claim 9, wherein the trained machine-learning model being trained includes splitting, by a computing device, the prior messages into the at least one normalized phrase extracted from the prior messages. ### Claim 13 (depends on 12) The non-transitory computer-readable medium of claim 12, wherein the at least one feature vector includes a first feature vector that includes at least one word associated with the prior messages. ### Claim 14 (depends on 13) The non-transitory computer-readable medium of claim 13, wherein applying the trained machine-learning model includes processing, by a computing system, the incoming message to generate a second feature vector based on the incoming message, wherein the second feature vector include the at least one word associated with the prior messages. ### Claim 15 (depends on 13) The non-transitory computer-readable medium of claim 13, wherein the first feature vector includes a count of the at least one word associated with the prior messages. ### Claim 16 A system comprising: a non-transitory computer-readable medium; and program instructions stored on the non-transitory computer-readable medium, the program instructions when executed by a processing device cause the processing device to: apply a trained machine-learning model to process an incoming message to a user account operating within a mobile computing device, the trained machine-learning model being trained using a training data set generated based on at least one normalized phrase extracted from prior messages sent from the user account and at least one feature vector associated with a set of the prior messages sent from the user account, responsive to applying the trained machine-learning model, determine a reply message by mapping the at least one normalized phrase extracted from the prior messages with the at least one feature vector, the reply message including at least one of the at least one normalized phrase extracted from the prior messages, generate from the reply message a relevant reply message to the incoming message, and update a graphic display of the mobile computing device with the reply message based on how often the reply message is selected or provided for selection. ### Claim 17 (depends on 16) The system of claim 16, comprising ranking the relevant reply message and generating a listing of the relevant reply message for display based on the ranking. ### Claim 18 (depends on 16) The system of claim 16, comprising receiving data input indicating selection of the relevant reply message and, responsively, sending the relevant reply message to another user account in a messaging conversation with the user account. ### Claim 19 (depends on 16) The system of claim 16, wherein the trained machine-learning model being trained includes splitting, by a computing device, the prior messages into the at least one normalized phrase extracted from the prior messages. ### Claim 20 (depends on 19) The system of claim 19, wherein the at least one feature vector includes a first feature vector that includes at least one word associated with the prior messages.