Management of speech technology modules in an interactive voice response system
First Claim
1. An interactive voice response (IVR) system having:
- a plurality of speech technology software modules, at least one of the modules comprising a speech recognition engine comprising a plurality of different language models, each different language model corresponding to a different type of voice channel for recognizing speech input from a user, at least one other of the modules comprising a text-to-speech engine for generating speech output for a user, and at least one other of the modules comprising a tone-recognition module for recognizing a plurality of pre-selected tones;
a voice application defining a plurality of interactions between a caller and the speech technology modules, each interaction having a task property and an interaction environment property; and
a speech technology selection module for selecting, for each interaction, one of the speech technology modules from the plurality of the modules to be used by the application according to the environment property of the interaction;
wherein the environment property comprises a call-type identifier that indicates when a call is a landline type call, a mobile type call, and an IP telephony type call, and wherein one of one of the plurality of different language models is used by the speech recognition engine depending on the call-type identifier.
2 Assignments
0 Petitions
Accused Products
Abstract
This invention relates to the management, in an interactive voice response system, of a plurality of speech technology modules. In particular it relates to an apparatus and a method for dynamically determining which of a plurality of speech technology modules to use during voice interaction between the system and a user. In prior art IVR systems each speech technology module is configured for a specific application or task. Most speech technology modules have different lexicons for the range of functions but it is the full lexicon which can determine an engine'"'"'s suitability for a language. For instance, one type of speech recognition engine is preferred for certain languages whereas IBM ViaVoice is a good general all rounder. Choosing one speech recognition module according to application or function alone is not entirely satisfactory and there is a need for improvement. The present solution is to select, for each interaction, one of the speech technology modules from the plurality of the modules to be used by the application according to the environment property of the interaction.
75 Citations
5 Claims
-
1. An interactive voice response (IVR) system having:
-
a plurality of speech technology software modules, at least one of the modules comprising a speech recognition engine comprising a plurality of different language models, each different language model corresponding to a different type of voice channel for recognizing speech input from a user, at least one other of the modules comprising a text-to-speech engine for generating speech output for a user, and at least one other of the modules comprising a tone-recognition module for recognizing a plurality of pre-selected tones; a voice application defining a plurality of interactions between a caller and the speech technology modules, each interaction having a task property and an interaction environment property; and a speech technology selection module for selecting, for each interaction, one of the speech technology modules from the plurality of the modules to be used by the application according to the environment property of the interaction; wherein the environment property comprises a call-type identifier that indicates when a call is a landline type call, a mobile type call, and an IP telephony type call, and wherein one of one of the plurality of different language models is used by the speech recognition engine depending on the call-type identifier. - View Dependent Claims (2, 3, 4, 5)
-
Specification