Process and system for semantically recognizing, correcting, and suggesting domain specific speech
First Claim
1. A method for processing a spoken utterance, comprising:
- (a) obtaining an input utterance, and(b) performing preliminary recognition using a speech engine and,(c) determining if one of the hypotheses from the N-best list of said speech engine is an exact string match to an entry in a database of sentences from a knowledge domain and,(d) marking said matching hypothesis from said knowledge domain as validated, or failing to identify a hypothesis in said knowledge domain marking all said hypotheses as invalid.
2 Assignments
0 Petitions
Accused Products
Abstract
Semantic understanding of hypotheses returned by a speech engine could improve the quality of recognition and in cases of misrecognition speed the identification of errors and potential substitutions. Unfortunately, semantic recognition using natural language parsers is hard since semantic and syntactic rules for processing language are complex and computationally expensive. Additionally, semantic recognition should be performed in a knowledge domain—a domain of interest such as radiology, pathology, or tort law. This adds additional complexity to building semantic rules. The method described achieves semantic understanding by coupling a speech recognition engine to a semantic recognizer, which draws from a database of domain sentences derived from a document corpus, and a knowledge base created for these domain sentences. The method is able to identify in near real-time the best sentence hypotheses from the speech recognizer and its associated meanings, i.e. propositions, which belong to the knowledge domain, and if no hypotheses are found mark it as invalid for easy identification. For invalid sentences, closely matched sentences from the domain database can be displayed to speed correction. The semantic propositions can be semantically typed, such as normal and abnormal, further aiding in the identification of recognition errors. The system automatically displays the semantic meaning of related ideas of valid sentences by retrieving proposition (s) from the knowledge base.
106 Citations
13 Claims
-
1. A method for processing a spoken utterance, comprising:
-
(a) obtaining an input utterance, and (b) performing preliminary recognition using a speech engine and, (c) determining if one of the hypotheses from the N-best list of said speech engine is an exact string match to an entry in a database of sentences from a knowledge domain and, (d) marking said matching hypothesis from said knowledge domain as validated, or failing to identify a hypothesis in said knowledge domain marking all said hypotheses as invalid. - View Dependent Claims (2, 3, 4, 12, 13)
-
- 5. A method for processing a valid hypothesis from a speech engine, consisting of obtaining matching proposition(s) from a semantic knowledge base using a semantic mapping table and displaying to the user said proposition(s) or failing to identify any propositions, marking the hypothesis as semantically unrecognized for a particular knowledge domain.
-
7. A method for obtaining related proposition(s) to a validated sentence, comprising:
-
(a) consisting of obtaining matching proposition(s) from a semantic knowledge base using a semantic mapping table and, (b) looking up related compound sentences from said semantic mapping table which match one or more of said matching propositions and, (c) returning as related proposition(s) from, said semantic mapping table which match said compound sentences. - View Dependent Claims (8, 9)
-
- 10. A method to retrieve from a database of domain sentences a list of closely matched sentences to an invalid sentence hypothesis produced by a speech engine.
Specification