Methods for understanding incomplete natural language query
First Claim
1. A system comprising:
- at least one processor; and
memory encoding computer executable instructions that, when executed by at least one processor, perform a method for interpreting incomplete natural language expressions, the method comprising;
receiving input from a client device, the user input comprising an incomplete natural language expression, wherein the incomplete natural language expression is a string of characters with one or more missing characters;
extracting one or more n-grams from the incomplete natural language expression;
analyzing the extracted one or more n-grams to determine a set of possible domains;
assigning a confidence level to each possible domain of the set of possible domains, wherein the confidence level for at least one possible domain exceeds a predetermined confidence threshold;
predicting an intent of a user associated with the at least one possible domain; and
initiating at least one domain application for performing the predicted intent;
predicting at least one slot for executing at least one function;
receiving a selection, wherein the received selection indicates confirmation that the at least one function reflects an actual intent of the user; and
training a natural language analysis component based on at least the received selection indicating confirmation.
2 Assignments
0 Petitions
Accused Products
Abstract
Analysis of incomplete natural language expressions using n-gram analysis and contextual information allows one or more domains to be predicted. For each domain, intent a likely intent of the user is determined using n-gram analysis and contextual information. Intent may correspond to functions of a domain application. In such a case, information required for the functions to execute the application may be populated using n-gram analysis and/or contextual information. The application may then be presented to the user for confirmation of intent. Confirmation of intent along with the incomplete natural language expression and contextual information may then be used to train one or more models used to predict user intent based on incomplete natural language expressions.
27 Citations
19 Claims
-
1. A system comprising:
-
at least one processor; and memory encoding computer executable instructions that, when executed by at least one processor, perform a method for interpreting incomplete natural language expressions, the method comprising; receiving input from a client device, the user input comprising an incomplete natural language expression, wherein the incomplete natural language expression is a string of characters with one or more missing characters; extracting one or more n-grams from the incomplete natural language expression; analyzing the extracted one or more n-grams to determine a set of possible domains; assigning a confidence level to each possible domain of the set of possible domains, wherein the confidence level for at least one possible domain exceeds a predetermined confidence threshold; predicting an intent of a user associated with the at least one possible domain; and initiating at least one domain application for performing the predicted intent; predicting at least one slot for executing at least one function; receiving a selection, wherein the received selection indicates confirmation that the at least one function reflects an actual intent of the user; and training a natural language analysis component based on at least the received selection indicating confirmation. - View Dependent Claims (2, 3, 4, 5, 12)
-
-
6. A computer implemented method, comprising:
-
receiving input from a client device, the user input comprising an incomplete natural language expression, wherein the incomplete natural language expression is a string of characters with one or more missing characters; extracting one or more n-grams from the incomplete natural language expression; analyzing the extracted one or more n-grams to determine a set of possible domains; assigning a confidence level to each possible domain of the set of possible domains, wherein the confidence level for at least one possible domain exceeds a predetermined confidence threshold; predicting an intent of a user associated with the at least one possible domain; initiating at least one domain application for performing the predicted intent; predicting at least one slot for executing at least one function; and receiving a selection, wherein the received selection indicates confirmation that the at least one function reflects an actual intent of the user; training a natural language analysis component based on at least the received selection indicating confirmation and the contextual information. - View Dependent Claims (7, 8, 9, 10, 11)
-
-
13. A computer readable storage device storing instructions that when executed perform the method of:
-
receiving textual input from a client device, the user input comprising an incomplete natural language expression, wherein the incomplete natural language expression is a string of characters with one or more missing characters; extracting one or more n-grams from the incomplete natural language expression; analyzing the extracted one or more n-grams to determine a first set of possible domains; assigning a confidence level to each possible domain of the first set of possible domains, wherein the confidence level for at least one possible domain exceeds a predetermined confidence threshold; predicting an intent of a user associated with the at least one possible domain; initiating at least one domain application for performing the predicted intent; predicting at least one slot for executing at least one function; and receiving a selection, wherein the received selection indicates confirmation that the at least one function reflects an actual intent of the user; training a natural language analysis component based on at least the received selection indicating confirmation. - View Dependent Claims (14, 15, 16, 17, 18, 19)
-
Specification