# US 11,443,112 B2 · Outcome of a natural language interaction > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,443,112 B2 - Document ID: US11443112B2 - Status: Granted - Kind: B2 - Application number: 16/563,270 - Filed: 2019-09-06 - Published: 2021-03-11 - Granted: 2022-09-13 - Earliest claimed: 2019-09-06 - Assignees: International Business Machines Corporation - Applicants: International Business Machines Corporation - Inventors: Jonathan F. Brunn; Rachael Marie Huston Dickens; Rui Zhang - Web page: https://patentgrep.com/patent/US11443112B2 ## CPC - G06F 40/30: Semantic analysis - G06F 17/18: for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)} - G06F 18/2415: based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate - G06F 18/2431: Multiple classes - G06F 40/205: Parsing - G06F 40/279: Recognition of textual entities - G06F 40/35: Discourse or dialogue representation - G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N) ## Abstract Using a natural language analysis, a current message is classified into a current message class, the current message being a portion of an interaction in narrative text form. For the interaction using a state prediction model, an interaction outcome corresponding to the current message class is forecasted, the forecasting comprising computing a probability that the current message class will result in a successful message class. Using the state prediction model, a set of next message classes and a set of predicted interaction outcomes are determined, each message in the set of next message classes corresponding to the current message class, each predicted interaction outcome in the set of predicted interaction outcomes corresponding to a next message class in the set of next message classes. According to the corresponding predicted interaction outcome, the set of next message classes is ranked. ## Claims ### Claim 1 A computer-implemented method comprising: training a state prediction model using a plurality of message sequences, each message in a message sequence classified into a message class, the training comprising configuring a Markov inference model to determine a plurality of conditional probabilities, each conditional probability comprising a probability of a message in a second message class following a message in a first message class, wherein a message class is represented by a node in the Markov inference model, the conditional probability is represented by a numerical value assigned to a link between corresponding nodes in the Markov inference model, wherein the training results in a trained state prediction model; classifying, using a natural language analysis, a current message into a current message class, the current message being a portion of an interaction in narrative text form; forecasting, for the interaction using the trained state prediction model, an interaction outcome corresponding to the current message class, the forecasting comprising computing a probability that the current message class will result in a successful message class, wherein the probability that the current message class will result in a successful message class is computed by evaluating a plurality of conditional probabilities corresponding to links connecting nodes in a plurality of paths through the Markov inference model; determining, using the trained state prediction model, a set of next message classes and a set of predicted interaction outcomes, each message in the set of next message classes corresponding to the current message class, each predicted interaction outcome in the set of predicted interaction outcomes corresponding to a next message class in the set of next message classes, wherein the set of next message classes and the set of predicted interaction outcomes are determined by evaluating the plurality of conditional probabilities in the plurality of paths through the Markov inference model, wherein a predicted interaction outcome in the set of predicted interaction outcomes is determined responsive to a conditional probability of a next message class in the set of next message classes being above a threshold value; and ranking, according to the corresponding predicted interaction outcome, the set of next message classes. ### Claim 2 (depends on 1) The computer-implemented method of claim 1, wherein the set of next message classes is determined according to a conditional probability of a next message class in the set of next message classes having a parent in the current message class. ### Claim 3 (depends on 2) The computer-implemented method of claim 2, wherein forecasting an interaction outcome corresponding to the current message class comprises: computing, for the current message class, a first probability of success, the first probability of success comprising a sum of a probability of success for each next message class in the set of next message classes multiplied by the conditional probability of the next message class in the set of next message classes having a parent in the current message class, the probability of success comprising a probability that a next message class will result in a successful message class. ### Claim 4 (depends on 1) The computer-implemented method of claim 1, further comprising: recommending a natural language response type, the natural language response type corresponding to a next message class in the set of next message classes, the natural language response type comprising a suggested response type to the current message intended to improve an interaction outcome corresponding to the next message class. ### Claim 5 (depends on 1) The computer-implemented method of claim 1, further comprising: constructing, using a natural language message construction model, a set of natural language responses, each natural language response in the set of natural language responses corresponding to a next message class in the set of next message classes, each natural language response in the set of natural language responses comprising a suggested response to the current message intended to improve an interaction outcome corresponding to the next message class. ### Claim 6 (depends on 1) The computer-implemented method of claim 1, further comprising: receiving a message identified as a parent message of the current message in the interaction; classifying, using the natural language analysis, the parent message into a parent message class; and updating, by updating the probability of the current message class having a parent in the parent message class, the trained state prediction model. ### Claim 7 A computer usable program product comprising one or more computer-readable storage devices, and program instructions stored on at least one of the one or more storage devices, the stored program instructions comprising: program instructions to train a state prediction model using a plurality of message sequences, each message in a message sequence classified into a message class, the training comprising configuring a Markov inference model to determine a plurality of conditional probabilities, each conditional probability comprising a probability of a message in a second message class following a message in a first message class, wherein a message class is represented by a node in the Markov inference model, the conditional probability is represented by a numerical value assigned to a link between corresponding nodes in the Markov inference model, wherein the training results in a trained state prediction model; program instructions to classify, using a natural language analysis, a current message into a current message class, the current message being a portion of an interaction in narrative text form; program instructions to forecast, for the interaction using the trained state prediction model, an interaction outcome corresponding to the current message class, the forecasting comprising computing a probability that the current message class will result in a successful message class, wherein the probability that the current message class will result in a successful message class is computed by evaluating a plurality of conditional probabilities corresponding to links connecting nodes in a plurality of paths through the Markov inference model; program instructions to determine, using the trained state prediction model, a set of next message classes and a set of predicted interaction outcomes, each message in the set of next message classes corresponding to the current message class, each predicted interaction outcome in the set of predicted interaction outcomes corresponding to a next message class in the set of next message classes, wherein the set of next message classes and the set of predicted interaction outcomes are determined by evaluating the plurality of conditional probabilities in the plurality of paths through the Markov inference model, wherein a predicted interaction outcome in the set of predicted interaction outcomes is determined responsive to a conditional probability of a next message class in the set of next message classes being above a threshold value; and program instructions to rank, according to the corresponding predicted interaction outcome, the set of next message classes. ### Claim 8 (depends on 7) The computer usable program product of claim 7, wherein the set of next message classes is determined according to a conditional probability of a next message class in the set of next message classes having a parent in the current message class. ### Claim 9 (depends on 8) The computer usable program product of claim 8, wherein program instructions to forecast an interaction outcome corresponding to the current message class comprises: program instructions to compute, for the current message class, a first probability of success, the first probability of success comprising a sum of a probability of success for each next message class in the set of next message classes multiplied by the conditional probability of the next message class in the set of next message classes having a parent in the current message class, the probability of success comprising a probability that a next message class will result in a successful message class. ### Claim 10 (depends on 7) The computer usable program product of claim 7, further comprising: program instructions to recommend a natural language response type, the natural language response type corresponding to a next message class in the set of next message classes, the natural language response type comprising a suggested response type to the current message intended to improve an interaction outcome corresponding to the next message class. ### Claim 11 (depends on 7) The computer usable program product of claim 7, further comprising: program instructions to construct, using a natural language message construction model, a set of natural language responses, each natural language response in the set of natural language responses corresponding to a next message class in the set of next message classes, each natural language response in the set of natural language responses comprising a suggested response to the current message intended to improve an interaction outcome corresponding to the next message class. ### Claim 12 (depends on 7) The computer usable program product of claim 7, further comprising: program instructions to receive a message identified as a parent message of the current message in the interaction; program instructions to classify, using the natural language analysis, the parent message into a parent message class; and program instructions to update, by updating the probability of the current message class having a parent in the parent message class, the state prediction model. ### Claim 13 (depends on 7) The computer usable program product of claim 7, wherein the stored program instructions are stored in the at least one of the one or more storage devices of a local data processing system, and wherein the stored program instructions are transferred over a network from a remote data processing system. ### Claim 14 (depends on 7) The computer usable program product of claim 7, wherein the stored program instructions are stored in the at least one of the one or more storage devices of a server data processing system, and wherein the stored program instructions are downloaded over a network to a remote data processing system for use in a computer readable storage device associated with the remote data processing system. ### Claim 15 A computer system comprising one or more processors, one or more computer-readable memories, and one or more computer-readable storage devices, and program instructions stored on at least one of the one or more storage devices for execution by at least one of the one or more processors via at least one of the one or more memories, the stored program instructions comprising: program instructions to train a state prediction model using a plurality of message sequences, each message in a message sequence classified into a message class, the training comprising configuring a Markov inference model to determine a plurality of conditional probabilities, each conditional probability comprising a probability of a message in a second message class following a message in a first message class, wherein a message class is represented by a node in the Markov inference model, the conditional probability is represented by a numerical value assigned to a link between corresponding nodes in the Markov inference model, wherein the training results in a trained state prediction model; program instructions to classify, using a natural language analysis, a current message into a current message class, the current message being a portion of an interaction in narrative text form; program instructions to forecast, for the interaction using the trained state prediction model, an interaction outcome corresponding to the current message class, the forecasting comprising computing a probability that the current message class will result in a successful message class, wherein the probability that the current message class will result in a successful message class is computed by evaluating a plurality of conditional probabilities corresponding to links connecting nodes in a plurality of paths through the Markov inference model; program instructions to determine, using the trained state prediction model, a set of next message classes and a set of predicted interaction outcomes, each message in the set of next message classes corresponding to the current message class, each predicted interaction outcome in the set of predicted interaction outcomes corresponding to a next message class in the set of next message classes, wherein the set of next message classes and the set of predicted interaction outcomes are determined by evaluating the plurality of conditional probabilities in the plurality of paths through the Markov inference model, wherein a predicted interaction outcome in the set of predicted interaction outcomes is determined responsive to a conditional probability of a next message class in the set of next message classes being above a threshold value; and program instructions to rank, according to the corresponding predicted interaction outcome, the set of next message classes. ### Claim 16 (depends on 15) The computer system of claim 15, wherein the set of next message classes is determined according to a conditional probability of a next message class in the set of next message classes having a parent in the current message class. ### Claim 17 (depends on 16) The computer system of claim 16, wherein program instructions to forecast an interaction outcome corresponding to the current message class comprises: program instructions to compute, for the current message class, a first probability of success, the first probability of success comprising a sum of a probability of success for each next message class in the set of next message classes multiplied by the conditional probability of the next message class in the set of next message classes having a parent in the current message class, the probability of success comprising a probability that a next message class will result in a successful message class.