System for using statistical classifiers for spoken language understanding
First Claim
Patent Images
1. A computer-readable storage medium including instructions readable by a computer which, when implemented, provide a text classifier in a natural language interface that receives a natural language user input, the text classifier comprising:
- a feature extractor extracting a binary feature vector of features from a textual input indicative of the natural language user input;
a statistical classifier coupled to the feature extractor, receiving the feature vector of features, and outputting a selected class identifier, the selected class identifier identifying a task to be completed, associated with the textual input, based on the features in the feature vector, the statistical classifier comprising;
a plurality of statistical classification components each statistical classification component receiving the feature vector and outputting a respective class identifier based on the feature vector, each respective class identifier representing one of a plurality of tasks to be completed that can be identified using each statistical classification component; and
a class selector coupled to the plurality of statistical classification components and selecting one of the respective class identifiers as the selected class identifier identifying the task to be completed wherein the class selector comprises one of a group comprising;
a voting component that selects a class identifier that is output by more statistical classification components than other class identifiers; and
an additional statistical classifier that receives as an input the class identifiers output from the plurality of statistical classification components and that selects one of the class identifiers received in the input; and
a parser that receives the selected class identifier and the textual input and generates a semantic representation of the textual input based on the selected class identifier.
4 Assignments
0 Petitions
Accused Products
Abstract
The present invention involves using one or more statistical classifiers in order to perform task classification on natural language inputs. In another embodiment, the statistical classifiers can be used in conjunction with a rule-based classifier to perform task classification.
84 Citations
22 Claims
-
1. A computer-readable storage medium including instructions readable by a computer which, when implemented, provide a text classifier in a natural language interface that receives a natural language user input, the text classifier comprising:
-
a feature extractor extracting a binary feature vector of features from a textual input indicative of the natural language user input; a statistical classifier coupled to the feature extractor, receiving the feature vector of features, and outputting a selected class identifier, the selected class identifier identifying a task to be completed, associated with the textual input, based on the features in the feature vector, the statistical classifier comprising; a plurality of statistical classification components each statistical classification component receiving the feature vector and outputting a respective class identifier based on the feature vector, each respective class identifier representing one of a plurality of tasks to be completed that can be identified using each statistical classification component; and a class selector coupled to the plurality of statistical classification components and selecting one of the respective class identifiers as the selected class identifier identifying the task to be completed wherein the class selector comprises one of a group comprising; a voting component that selects a class identifier that is output by more statistical classification components than other class identifiers; and an additional statistical classifier that receives as an input the class identifiers output from the plurality of statistical classification components and that selects one of the class identifiers received in the input; and a parser that receives the selected class identifier and the textual input and generates a semantic representation of the textual input based on the selected class identifier. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17)
-
-
18. A computer-implemented method of processing a natural language input for use in completing a task represented by the natural language input, comprising:
-
generating a feature vector based on the natural language input; performing statistical classification on the feature vector with a processor to obtain a class identifier identifying a target class associated with the natural language input, the statistical classification being performed by performing statistical classification on the natural language input using a plurality of different statistical classifiers, and selecting the class identifier output by a greatest number of the plurality of statistical classifiers as representing the target class; selectively activating grammar rules in a grammar used in a rule-based analyzer, the activated grammar rules corresponding to the class identifier; and analyzing the natural language input with the rule-based analyzer using the grammar with the activated grammar rules to generate semantic expressions from the natural language input to fill semantic slots in the target class. - View Dependent Claims (19, 20, 21)
-
-
22. A computer-implemented method of processing a natural language input for use in completing a task represented by the natural language input, comprising:
-
generating a feature vector based on the natural language input; performing statistical classification on the feature vector with a processor to obtain a class identifier identifying a target class associated with the natural language input; selectively activating grammar rules in a grammar used in a rule-based analyzer, the activated grammar rules corresponding to the class identifier; analyzing the natural language input with the rule-based analyzer using the grammar with the activated grammar rules to generate semantic expressions from the natural language input to fill semantic slots in the target class; performing rule-based analysis on the natural language input to obtain another class identifier; and identifying the target class based on the class identifier obtained from the statistical classification and the another class identifier obtained from the rule-based analysis.
-
Specification