User barge-in enablement in large vocabulary speech recognition systems
First Claim
1. A method for communicating with a customer who communicates into a mouthpiece of telephone comprising the steps of:
- applying a signal received by said mouthpiece to a preprocessing module that discards from said signal those components of said signal that fail to meet preselected usefulness criteria threshold test determined based on task-related speech of the speaker, resulting in an intermediate signal;
applying said intermediate signal to a phrase detector, to detect meaningful phrases within said intermediate signal; and
supplying to a controller data that corresponds to detected meaningful phrases.
4 Assignments
0 Petitions
Accused Products
Abstract
An interactive voice response unit which provides beneficial operation by including means to handle unconstrained input such as natural speech and to allow barge-in includes a prompter, a recognizer of speech signals, a meaningful phrase detector and classifier, and a turn-taking module, all under control of a dialog manager. In the course of listening to user input while outputting a voiced message, the voice response unit processes the received signal and ascertains whether it is receiving an utterance that is intended to interrupt the prompt, or merely noise or an utterance that is not meant to be used by the arrangement. The unit is sensitive to the speed and context of the speech provided by the user and is thus able to distinguish between a situation where a speaker is merely pausing and a situation where a speaker is done speaking.
79 Citations
25 Claims
-
1. A method for communicating with a customer who communicates into a mouthpiece of telephone comprising the steps of:
-
applying a signal received by said mouthpiece to a preprocessing module that discards from said signal those components of said signal that fail to meet preselected usefulness criteria threshold test determined based on task-related speech of the speaker, resulting in an intermediate signal;
applying said intermediate signal to a phrase detector, to detect meaningful phrases within said intermediate signal; and
supplying to a controller data that corresponds to detected meaningful phrases. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8)
-
-
9. A method for communicating with a customer who speaks to a mouthpiece of a telephone comprising the steps of:
-
analyzing a signal received by said mouthpiece to determine whether said signal meets preselected usefulness criteria threshold test determined based on task-related speech of the speaker;
applying to a phrase detector those components of said signal that meet said usefulness criteria, to detect meaningful phrases within said signal; and
supplying to a controller data that corresponds to said meaningful phrases. - View Dependent Claims (10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25)
said controller determining that an action is to be taken, even before a silence period is detected in said incoming signal, and said controller taking such action even before said silence period is detected in said incoming signal.
-
-
15. The method of claim 14 where said action is a prompt signal delivered to said customer.
-
16. The method of claim 14 where said action is a cutting off of a prompt signal delivered to said customer.
-
17. The method of claim 14 where said action is a stopping of a prompt signal delivered to customer.
-
18. The method of claim 9 where said processing step that removes influence is an echo canceling processing step.
-
19. The method of claim 9 where said phrase detector analyzes a rate at which words are detected, and lengths of silences between detected words, to control a determination that a response is completed.
-
20. The method of claim 9 where said phrase detector detects significant speech phrases through word spotting.
-
21. The method of claim 9 where said phrase detector analyzes inflection of detected words.
-
22. The method claim 9 where said phrase detector employs detected words and detected silences between words to ascertain likelihood that useful input in said incoming signal has ended.
-
23. The method claim 9 where said phrase detector carries out the steps of:
-
translating meaningful phrases found in said output signals of said recognizer into tasks to be provided to said controller, and ascertaining lengths of silences between significant signal segments detected by said phrase detector to ascertain likelihood that useful input in said incoming signal has ended.
-
-
24. The method claim 9 where said phrase detector carries out the step of determining, from grammatical construct detected phrases, whether said customer has completed sending information.
-
25. The method of claim 9 where said signal received by said mouthpiece comprises one or more components from a set containing elements a, b, c, and d, where
element a is a signal component that results from sounds made by said customer and intended to communicate information to said mouthpiece, element b is a signal component that results from sounds made by said customer that is not intended to communicate information to said mouthpiece, element c is a signal component that results from non-speech sounds made by said customer, and element d is a signal component that results from sounds made by other than said customer, and where element a is the only element that meets said usefulness criteria.
Specification