System and method for evaluating intent of a human partner to a dialogue between human user and computerized system
First Claim
Patent Images
1. A computerized method for assigning relative scores to various possible intents on the part of a user approaching a virtual agent, the method comprising:
- employing a processor for predicting priority topics, including gathering first data and employing the first data to discern and seek user confirmation of at least one possible intent on the part of the user; and
subsequent to receipt of said confirmation, employing a processor for gathering second data and employing the second data to provide service to the user to suit the user'"'"'s confirmed intent,wherein a context tree having nodes is used to identify all topics and entities that occurred in a multi-step dialogue up to a current dialogue state corresponding to a current dialogue step, and their relevance for a current context created by said current dialogue step and wherein relevance between said nodes is related to distance therebetween,wherein intent scoring occurs in conjunction with a hosting dialogue system operative to resolve external data dependencies and wherein the hosting dialogue system resolves at least a first item employing available user data to produce an intent prediction result, including assigning relevance scores to significant topic closures and wherein at least said intention prediction result is used in a subsequent item to display a guided natural additional item, prediction scores are used to enhance NLP results and a most suitable topic is selected as a next focus for discussion with said user, andwherein dialog is conducted accordingly, with a mobile communication device serving the user.
3 Assignments
0 Petitions
Accused Products
Abstract
A system and method for assigning relative scores to various possible intents on the part of a user approaching a virtual agent, the method comprising predicting priority topics, including gathering first data and employing the first data to discern and seek user confirmation of at least one possible intent on the part of the user; and subsequent to receipt of the confirmation, gathering second data and employing the second data to provide service to the user to suit the user'"'"'s confirmed intent.
48 Citations
24 Claims
-
1. A computerized method for assigning relative scores to various possible intents on the part of a user approaching a virtual agent, the method comprising:
-
employing a processor for predicting priority topics, including gathering first data and employing the first data to discern and seek user confirmation of at least one possible intent on the part of the user; and subsequent to receipt of said confirmation, employing a processor for gathering second data and employing the second data to provide service to the user to suit the user'"'"'s confirmed intent, wherein a context tree having nodes is used to identify all topics and entities that occurred in a multi-step dialogue up to a current dialogue state corresponding to a current dialogue step, and their relevance for a current context created by said current dialogue step and wherein relevance between said nodes is related to distance therebetween, wherein intent scoring occurs in conjunction with a hosting dialogue system operative to resolve external data dependencies and wherein the hosting dialogue system resolves at least a first item employing available user data to produce an intent prediction result, including assigning relevance scores to significant topic closures and wherein at least said intention prediction result is used in a subsequent item to display a guided natural additional item, prediction scores are used to enhance NLP results and a most suitable topic is selected as a next focus for discussion with said user, and wherein dialog is conducted accordingly, with a mobile communication device serving the user. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17)
-
-
18. A system for assigning relative scores to various possible intents on the part of a user approaching a virtual agent, the system comprising:
-
a dialog topic predictor operative for predicting priority topics, including gathering first data and employing the first data to discern and seek user confirmation of at least one possible intent on the part of the user; and a dialog generator operative, subsequent to receipt of said confirmation, for gathering second data and employing the second data to provide dialog with a user to suit the user'"'"'s confirmed intent, wherein at least one predictor vector is employed to increase accuracy of at least one mid-dialogue fork prediction process used to generate a relevance vector for at least one available fork topic, including modifying probabilities assigned by the fork prediction process thereby to generate improved predictions and using said improved predictions for at least one interface. - View Dependent Claims (19, 20, 21, 22, 23)
-
-
24. A computer program product, comprising a non-transitory tangible computer-readable medium having computer-readable program code embodied therein, said computer-readable program code adapted to be executed to implement a method for assigning relative scores to various possible intents on the part of a user approaching a virtual agent, the method comprising:
-
predicting priority topics, including gathering first data and employing the first data to discern and seek user confirmation of at least one possible intent on the part of the user; and subsequent to receipt of said confirmation, gathering second data and employing the second data to provide service to the user to suit the user'"'"'s confirmed intent, wherein a context tree having nodes is used to identify all topics and entities that occurred in a multi-step dialogue up to a current dialogue state corresponding to a current dialogue step, and their relevance for a current context created by said current dialogue step and wherein relevance between said nodes is inversely related to distance therebetween, wherein variable binding is used to automatically generate at least one appropriate natural language name for at least one topic closure, and wherein said binding is generated using Rete rules which identify patterns in a Cartesian multiplication between topic from topic ontology and relevant data for an open variable from an external system.
-
Specification