Back to results
US 11,888,792 B2Granted
Bot supervision
AssigneeLiveperson, Inc.
InventorsMichael Higgins, Xi Chen, Chuqinh He, Ender Ricart
Life of this patent
Estimating the term…
Abstract
The present disclosure relates generally to facilitating routing of communications. More specifically, techniques are provided to dynamically routing messages having multiple intents between bots and terminal devices during communication sessions configured with multi-channel capabilities.
Claims30
A computer-implemented method comprising:
- storing a set of supervision criteria in memory, the set of supervision criteria including a threshold level of confidence;
- monitoring one or more messages in a conversation between a bot and a user of a client device in accordance with the set of supervision criteria;
- detecting that a bot prediction associated with the conversation fails to meet the threshold level of confidence associated with the set of supervision criteria;
- requesting input from an agent device based on the bot prediction failing to meet the threshold level of confidence, wherein the agent device provides annotation input identifying one or more of the messages in the conversation and identifying contextual or supplemental information related to classification of the one or more identified messages in the conversation;
- and
- updating a learning model based on the one or more messages identified by the annotation input and the classification according to the identified contextual or supplemental information, wherein the bot uses the updated learning model to make a next bot prediction regarding a next conversation having one or more similar messages.
3 more elements
A system comprising:
- memory that stores a set of supervision criteria in memory, the set of supervision criteria including a threshold level of confidence;
- a communication interface that receives one or more messages in a conversation between a bot and a user of a client device in accordance with the set of supervision criteria;
- and
- a processor that executes instructions stored in memory, wherein the processor executes the instructions to:
- monitor the received messages in the conversation;
- detect that a bot prediction associated with the conversation fails to meet the threshold level of confidence associated with the set of supervision criteria;
- request input from an agent device based on the bot prediction failing to meet the threshold level of confidence, wherein the agent device provides annotation input identifying one or more of the messages in the conversation and identifying contextual or supplemental information related to classification of the one or more identified messages in the conversation;
- and
- update a learning model based on the one or more messages identified by the annotation input and the classification according to the identified contextual or supplemental information, wherein the bot uses the updated learning model to make a next bot prediction regarding a next conversation having one or more similar messages.
6 more elements
A non-transitory machine-readable storage medium, including instructions that, when executed by one or more processors, cause the one or more processors to perform operations including:
- storing a set of supervision criteria in memory, the set of supervision criteria including a threshold level of confidence;
- monitoring one or more messages in a conversation between a bot and a user of a client device in accordance with the set of supervision criteria;
- detecting that a bot prediction associated with the conversation fails to meet the threshold level of confidence associated with the set of supervision criteria;
- requesting input from an agent device based on the bot prediction failing to meet the threshold level of confidence, wherein the agent device provides annotation input identifying one or more of the messages in the conversation and identifying contextual or supplemental information related to classification of the one or more identified messages in the conversation;
- and
- updating a learning model based on the one or more messages identified by the annotation input and the classification according to the identified contextual or supplemental information, wherein the bot uses the updated learning model to make a next bot prediction regarding a next conversation having one or more similar messages.
3 more elements
Dependencies are inferred from the claim text.