Abstract: A system which uses automatic speech recognition to provide dialogs with human speakers automatically detects one or more characteristics, which may be characteristics of a speaker, his speech, his environment, or the speech channel used to communicate with the speaker. The characteristic may be detected either during the dialog or at a later time based on stored data representing the dialog. If the characteristic is detected during the dialog, the dialog can be customized for the speaker at an application level, based on the detected characteristic. The customization may include customization of operations and features such as call routing, error recovery, call flow, content selection, system prompts, or system persona. Data indicative of detected characteristics can be stored and accumulated for many speakers and/or dialogs and analyzed offline to generate a demographic or other type of analysis of the speakers or dialogs with respect to one or more detected characteristics.

Abstract: A computerized method for building and running natural language understanding systems, wherein a natural language understanding system takes a sentence as input and returns some representation of the possible meanings of the sentence as output (the “interpretation”) using a run-time interpreter that assigns interpretations to sentences and a compiler that produces (in a computer memory) an internal specification needed for the run-time interpreter from a user specification of the semantics of the application. The compiler builds a natural language system, while the run-time interpreter runs the system.

Abstract: A method and apparatus are used for analyzing dialogs in an interactive natural language speech recognition system. As a telephone call is received in the interactive natural language speech recognition system, user utterances are digitized and stored. A data log is formed which stores the prompt played, the dialog state, the address location of the audio recording of the user's responses, the recognition grammar, the recognition result type, the recognized word string, the understood meaning/interpretation, and the recognition latency. The telephone call can be virtually reconstructed from the stored user utterances and the data log. An automatic data analysis tool provides information regarding the relative success and failure of the various dialogs as an aid to modifying the dialogs to improve system performance.

Abstract: An automatic speech recognition methodology, wherein words are modeled as probabilistic networks of allophones, collects nodes in the probabilistic network into equivalence classes when those nodes have the same allophonic choices governed by the same phonological rules. The allophonic choices allow for representation of dialectic pronunciation variations between different speakers. Training data is shared among nodes in an equivalence class so that accurate pronunciation probabilities may be determined even for words for which there is only a limited amount of training data. A method is used to determine probabilities for each of a multitude of pronunciation models for each word in the vocabulary, based on automatic extraction of linguistic knowledge from sets of phonological rules, in order to robustly and accurately model dialectal variation.

Abstract: In a hidden Markov model-based speech recognition system, multilayer perceptrons (MLPs) are used in context-dependent estimation of a plurality of state-dependent observation probability distributions of phonetic classes. Estimation is obtained by the Bayesian factorization of the observation likelihood in terms of posterior probabilities of phone classes assuming the context and the input speech vector. The context-dependent estimation is employed as the state-dependent observation probabilities needed as parameter input to a hidden Markov model speech processor to identify the word sequence representing the unknown speech input of input speech vectors. Within the speech processor, models are provided which employ the observation probabilities in the recognition process.

Abstract: An automatic speech recognition methodology takes advantage of linguistic constraints wherein words are modeled as probabilistic networks of phonetic segments (herein phones), and each phone is represented as a context-independent hidden Markov phone model mixed with a number of context-dependent phone models. Recognition is based on use of methods to design phonological rule sets based on measures of coverage and overgeneration of pronunciations which achieves high coverage of pronunciations with compact representations. Further, a method estimates probabilities of the different possible pronunciations of words. A further method models cross-word coarticulatory effects. In a specific embodiment of the system, a specific method determines the single most-likely pronunciation of words. In further specific embodiments of the system, methods generate speaker-dependent pronunciation networks.