A voice input device provided with an input section for inputting a voice of a user, a recognition section for recognizing the voice of the user inputted by the input section, a generation section for generating characters or a command based on a recognition result of the recognition section, a detection section for detecting a device's own posture, and an instruction section for instructing the generation section to generate the command when a detection result of the detection section represents a specific posture as compared to instructing the generation section to generate the characters when the detection result of the detection section represents a posture other than the specific posture. Accordingly, character input and command input during dictation is correctly distinguished, or more specifically unexpected character input during dictation is avoided.

Methods, apparatus, systems and articles of manufacture are disclosed to determine the speed-up of media programs using speech recognition. An example apparatus disclosed herein is to perform speech recognition on a first audio clip collected by a media meter to recognize a first text string associated with the first audio clip, compare the first text string to a plurality of reference text strings associated with a corresponding plurality of reference audio clips to identify a matched one of the reference text strings, and estimate a presentation rate of the first audio clip based on a first time associated with the first audio clip and a second time associated with a first one of the reference audio clips corresponding to the matched one of the reference text strings.

Methods of incrementally modifying a word-level finite state transducer (FST) are described for adding and removing sentences. A prefix subset of states and arcs in the FST is determined that matches a prefix portion of the sentence. A suffix subset of states and arcs in the FST is determined that matches a suffix portion of the sentence. A new sentence can then be added to the FST by appending a new sequence of states and arcs to the FST corresponding to a remainder of the sentence between the prefix and suffix. An existing sentence can be removed from the FST by removing any arcs and states between the prefix subset and the suffix subset. The resulting modified FST is locally efficient but does not satisfy global optimization criteria such as minimization.

A method for processing a voice command using a statistical dialog model determines a belief state as a probability distribution over states organized in a hierarchy with a parent-child relationship of nodes representing the states. The belief state includes the hierarchy of state variables defining probabilities of each state to correspond to the voice command and a probability of a state of a child node in the hierarchy is conditioned on a probability of a state of a corresponding parent node. A system action is selected based on the belief state.

A voice to text model used by a voice-enabled electronic device is dynamically and in a context-sensitive manner updated to facilitate recognition of entities that potentially may be spoken by a user in a voice input directed to the voice-enabled electronic device. The dynamic update to the voice to text model may be performed, for example, based upon processing of a first portion of a voice input, e.g., based upon detection of a particular type of voice action, and may be targeted to facilitate the recognition of entities that may occur in a later portion of the same voice input, e.g., entities that are particularly relevant to one or more parameters associated with a detected type of voice action.

Embodiments describe a method for decoding speech including receiving speech input at an audio input device, generating speech data that is a digital representation of the speech input; extracting acoustic features of the speech data, assigning acoustic scores to the acoustic features, receiving data representing the acoustic features and the acoustic scores, decoding the data representing the acoustic features into a word, having a word score, by referencing a WFST language model, modifying the word score into a new word score based on a personalized grammar model stored in the external memory device, the processor is separate from and external to the WFST accelerator, and determining an intent represented by a plurality of words outputted by the WFST accelerator, where the plurality of words include the word and the new word score.

A method includes generating an audio signal encoding an utterance captured by a microphone of a user device and transmitting the audio signal encoding the utterance to a server. The server is configured to determine a speaker of the utterance from one of a plurality of different users of the user device based on a comparison between the audio signal encoding the utterance and corresponding speaker verification data, and process the audio signal encoding the utterance using a speech recognition module to identify a particular action. The method also includes executing the particular action identified by the server to cause a particular application to launch on the user device based on user permissions associated with the speaker determined by the server to access the particular data.

This document relates to machine learning. One example includes a method or technique that can be performed on a computing device. The method or technique can include obtaining a task-adapted generative model that has been tuned using one or more task-specific seed examples. The method or technique can also include inputting dialog acts into the task-adapted generative model and obtaining synthetic utterances that are output by the task-adapted generative model. The method or technique can also include populating a synthetic training corpus with synthetic training examples that include the synthetic utterances. The synthetic training corpus may be suitable for training a natural language understanding model.

A method for routing audio content through an electronic device that is to be worn by a user. The method obtains a communication and determines whether the communication is private. In response to determining that the communication is private, the method drives a bone conduction transducer of the electronic device with an audio signal associated with the communication. In response to determining that the communication is not private, however, the method drives a speaker of the electronic device with the audio signal.

A method for streaming action fulfillment receives audio data corresponding to an utterance where the utterance includes a query to perform an action that requires performance of a sequence of sub-actions in order to fulfill the action. While receiving the audio data, but before receiving an end of speech condition, the method processes the audio data to generate intermediate automated speech recognition (ASR) results, performs partial query interpretation on the intermediate ASR results to determine whether the intermediate ASR results identify an application type needed to perform the action and, when the intermediate ASR results identify a particular application type, performs a first sub-action in the sequence of sub-actions by launching a first application to execute on the user device where the first application is associated with the particular application type. The method, in response to receiving an end of speech condition, fulfills performance of the action.