Abstract: The invention provides techniques for creating and using fragmented word models to increase the effective size of an active vocabulary of a speech recognition system. The active vocabulary represents all words and word fragments that the speech recognition system is able to recognize. Each word may be represented by a combination of acoustic models. As such, the active vocabulary represents the combinations of acoustic models that the speech recognition system may compare to a user's speech to identify acoustic models that best match the user's speech. The effective size of the active vocabulary may be increased by dividing words into constituent components or fragments (for example, prefixes, suffixes, separators, infixes, and roots) and including each component as a separate entry in the active vocabulary.

Abstract: The content of a speech sample is recognized using a computer system by evaluating the speech sample against a nonparametric set of training observations, for example, utterances from one or more human speakers. The content of the speech sample is recognized based on the evaluation results. The speech recognition process also may rely on a comparison between the speech sample and a parametric model of the training observations.

Abstract: Enrolling a user in a speech recognition system includes analyzing acoustic content of a user utterance and determining whether the utterance matches a portion of an enrollment text. If the user utterance matches a portion of the enrollment text, the acoustic content is used to update acoustic models corresponding to the portion of the enrollment text.

Abstract: Recognition of speech by a speech recognizer is improved by receiving deliberately contiguously repeated spoken utterances corresponding to a speech element and recognizing fewer instances of the speech element than the number of repeated spoken utterances. If a spoken utterance corresponding to the speech element is received and misrecognized prior to receiving the deliberately contiguously repeated spoken utterances, the spoken utterance and the repeated spoken utterances may be used to recognize the speech element.

Abstract: Language model results are combined according to a combination expression to produce combined language model results for a set of candidates. A candidate is selected and the combination expression is adjusted using language model results associated with the selected candidate.

Abstract: A computer enrolls a user in a speech recognition system by obtaining data representing a user's speech, the speech including multiple user utterances and generally corresponding to an enrollment text, and analyzing acoustic content of data corresponding to a user utterance. The computer determines, based on the analysis, whether the user utterance matches a portion of the enrollment text. If so, the computer uses the acoustic content of the user utterance to update acoustic models corresponding to the portion of the enrollment text. The computer may determine that the user utterance matches a portion of the enrollment text even when the user has skipped or repeated words of the enrollment text.

Abstract: A source-adapted model for use in speech recognition is generated by defining a linear relationship between a first element of an initial model and a first element of the source-adapted model. Thereafter, speech data that corresponds to the first element of the initial model is assembled from a set of speech data for a particular source associated with the source-adapted model. A linear transform that maps between the assembled speech data and the first element of the initial model is then determined. Finally, a first element of the source-adapted model is produced from the first element of the initial model using the linear transform.

Abstract: A speech sample is recognized with a computer system by processing the speech sample with at least two speech recognizers, each of which has a different performance characteristic. One speech recognizer may be a large-vocabulary, continuous speech recognizer optimized for real-time responsiveness and another speech recognizer may be an offline recognizer optimized for high accuracy. The speech content of the sample is recognized based on processing results from the speech recognizers. The speaker is provided with a real-time, yet potentially error-laden, text display corresponding to the speech sample while, subsequently, a human transcriptionist may use recognition results from multiple recognizers to produce an essentially error-free transcription. The performance characteristics of the recognizers may be based on style or subject matter, and the recognizers may operate serially or in parallel.

Abstract: A user searches searchable information representing underlying information using an interactive user interface, which indicates to the user which parts of the searchable information represent information in the underlying information that is of interest to the user.

Abstract: A word recognition system can: respond to the input of a character string from a user by limiting the words it will recognize to words having a related, but not necessarily the same, string; score signals generated after a user has been prompted to generate a given word against words other than the prompted word to determine if the signal should be used to train the prompted word; vary the number of signals a user is prompted to generate to train a given word as a function of how well the training signals score against each other or prior models for the prompted word; create a new acoustic model of a phrase by concatenating prior acoustic models of the words in the phrase; obtain information from another program running on the same computer, such as its commands or the context of text being entered into it, and use that information to vary which words it can recognize; determine which program unit, such as an application program or dialog box, currently has input focus on its computer and create a vocabulary

Abstract: A word recognition system can: respond to the input of a character string from a user by limiting the words it will recognize to words having a related, but not necessarily the same, string, score signals generated after a user has been prompted to generate a given word against words other than the prompted word to determine if the signal should be used to train the prompted word; vary the number of signals a user is prompted to generate to train a given word as a function of how well the training signals score against each other or prior models for the prompted word; create a new acoustic model of a phrase by concatenating prior acoustic models of the words in the phrase; obtain information from another program running on the same computer, such as its commands or the context of text being entered into it, and use that information to vary which words it can recognize; determine which program unit, such as an application program or dialog box, currently has input focus on its computer and create a vocabulary

Abstract: A method of adding a word to a speech recognition vocabulary includes creating a collection of possible phonetic pronunciations from a spelling of the word and using speech recognition to find a pronunciation from the collection that best matches an utterance of the word. The collection is created by comparing the spelling to a rules list of letter strings with associated phonemes. The list is searched for a letter string from the spelling of length greater than one letter. The collection is limited to phonetic pronunciations containing phonemes associated with the letter string of length greater than one. In another method, a net of possible phonetic pronunciations of the word is created from the spelling and speech recognition is used to find the pronunciation from the net that best matches the utterance of the word. The invention also features methods of assigning a pre-filtering class to a word.

Abstract: A word recognition system can: respond to the input of a character string from a user by limiting the words it will recognize to words having a related, but not necessarily the same, string; score signals generated after a user has been prompted to generate a given word against words other than the prompted word to determine if the signal should be used to train the prompted word; vary the number of signals a user is prompted to generate to train a given word as a function of how well the training signals score against each other or prior models for the prompted word; create a new acoustic model of a phrase by concatenating prior acoustic models of the words in the phrase; obtain information from another program running on the same computer, such as its commands or the context of text being entered into it, and use that information to vary which words it can recognize; determine which program unit, such as an application program or dialog box, currently has input focus on its computer and create a vocabulary

Abstract: Incorrect text associated with recognition errors in computer-implemented speech recognition is corrected by performing speech recognition on an utterance to produce a recognition result for the utterance. When a correction command is identified in the recognition result for the utterance, corrected text is produced from a portion of the recognition result for the utterance.

Abstract: System for segmenting text and identifying segment topics that match a user-specified topic. Topic tracking system creates a set of topic models from training text containing topic boundaries using a clustering algorithm. User supplies topic text. System creates a topic model of the topic text and adds the topic model to the set of topic models. User-supplied test text is segmented according to the set of topic models. Segments relating to the same topic as the topic text are selected.

Abstract: A speech sample is evaluated using a computer. Training data that include samples of speech are received and stored along with identification of speech elements to which portions of the training data are related. A speech sample is received and speech recognition is performed on the speech sample to produce recognition results. Finally, the recognition results are evaluated in view of the training data and the identification of the speech elements to which the portions of the training data are related. The technique may be used to perform tasks such as speech recognition, speaker identification, and language identification.

Abstract: A speech recognition system includes a speech-response capability for responding to sounds which appear to match models of spoken words by performing functions associated with such words. This speech response can be turned on or off. If the system detects both that speech response is off and that no indication is stored that the off state has been confirmed by a user, it performs a confirmation process. This prompts the user to utter a phrase confirming whether or not speech response is to be off; turns speech response on so it can respond to the user's confirmation utterance; determines whether to leave speech response on or off in response to the word which the speech response selects as corresponding to the user's confirmation utterance; and responds to such a determination that speech response is to be turned off, both by turning it off and by storing an indication that the off state has been confirmed by the user.

Abstract: A word recognition system is disclosed for converting spoken utterances into either text or commands. The system runs on a platform capable of running a plurality applications. Text and commands are sent from a word recognition application to one or more user applications. In addition, information pertaining to the state of the user applications is sent back to the word recognition application. Word recognition probabilities are modified based the information received from the user applications. As a result, the probabilities of recognizing a spoken utterance as a particular command will be greater when that command is active in the user application then when it is not active. Also, text strings will be assigned higher probabilities when they are appropriate for the present state of the user application.

Abstract: A speech model is produced for use in determining whether a speaker associated with the speech model produced an unidentified speech sample. First a sample of speech of a particular speaker is obtained. Next, the contents of the sample of speech are identified using speech recognition. Finally, a speech model associated with the particular speaker is produced using the sample of speech and the identified contents thereof. The speech model is produced without using an external mechanism to monitor the accuracy with which the contents were identified.

Abstract: A computerized word recognition system, such as a speech recognition system, stores word models of a first and second set for each of a plurality of vocabulary words. The system has a user interface which enable a user to selectively prevent the use of a second set's word model for a selected word. Often the first set of word models are spelled word models, such as models represented by a sequence of phonetic component models, each of is derived from similar speech sounds occurring in different words. In such systems the second set of word models are custom words models derived largely from word signals which are presumed to correspond only to the model's associated word. In many embodiments, the user interface allows a user to select to stop using a selected word's custom, or second set, model by selecting a menu or control window of a user interface.