Abstract: Techniques for remotely executing a secondary-device driver for generating commands for a secondary device are described herein. For instance, a secondary device (or “appliance”) may reside within an environment, along with a device to which the secondary device communicatively couples. The device may be configured to send control signals to the secondary device for causing the secondary device to perform certain operations. For instance, a user in the environment may provide, to the device, a request that the secondary device perform a certain operation. The device, which may lack some or all of a device driver associated with the secondary device, may then work with a remote service that executes the device driver for the purpose of receiving a command from the device driver and sending the command along to the secondary device. Upon receiving the command, the secondary device may perform the operation.

Abstract: Techniques for causing different devices to perform different operations using a single voice command are described herein. In some instances, a user may define a “scene”, in which a user sets different devices to different states and then associates an utterance with those states or with the operations performed by the devices to reach those states. For instance, a user may dim a light, turn on his television, and turn on his set-top box before sending a request to a local device or to a remote service to associate those settings with a predefined utterance, such as “my movie scene”. Thereafter, the user may cause the light to dim, the television to turn on, and the set-top box to turn on simply by issuing the voice command “execute my movie scene”.

Abstract: Aspects described herein may relate to the determination of data that is indicative of a greater range of speech properties than input text data. The determined data may be used as input to one or more speech processing tasks, such as model training, model validation, model testing, or classification. For example, after a model is trained based on the determined data, the model's performance may exhibit more resilience to a wider range of speech properties. The determined data may include one or more modified versions of the input text data. The one or more modified versions may be associated with the one or more speakers or accents and/or may be associated with one or more levels of semantic similarity in relation to the input text data. The one or more modified versions may be determined based on one or more machine learning algorithms.

Abstract: An electronic device includes a microphone configured to receive input of speech, an output interface, and a processor configured to recognize the speech inputted to the microphone. The processor judges whether the electronic device satisfies a predetermined condition. The processor causes the output interface to output information prompting a user for speech input when the processor judges that the electronic device satisfies the predetermined condition. This electronic device improves the convenience of speech recognition.

Abstract: The present technology relates to an information processing device, and an information processing method, each of which enables to reduce a confirmation load put on a user before a task is executed. The information processing device according to one embodiment of the present technology has the feature of, on the basis of relationship between a first cost required in a case where execution of a predetermined task has been a mistake and a second cost that is allowed by a user for the predetermined task that has been executed by mistake, calculating a confirmation degree of confirming the user as to whether or not to execute the predetermined task, and performing the confirmation by contents corresponding to the calculated degree. The present technology can be applied to an agent apparatus that operates using a voice UI.

Abstract: A processing system operates in a first power domain and includes a first memory, a memory access circuit, and a first processing circuit. The first memory stores sound data detected by a microphone. The memory access circuit transfers the sound data to a second memory according to a first command, in order to store the sound data as voice data. The first processing circuit outputs a second command according to a human voice detection signal. The second command is for enabling a second processing circuit, in order to determine whether the voice data in the second memory matches a predetermined voice command. One of the first and the second processing circuits outputs the first command. The second processing circuit operates in a second power domain. A power consumption to which the first power domain corresponds is lower than a power consumption to which the second power domain corresponds.

Abstract: An electronic record voice assistant system can include one or more processors that receive audio data, apply a machine learning model to the audio data to generate speech data including at least one value, determine a state of an electronic record, and update one or more fields of the electronic record using the state and the at least one value.

Abstract: Method and apparatus for speech processing are disclosed. A first unvoicing parameter for a first frame of a speech signal is determined, and furthered smoothed based on a second unvoicing parameter for a second frame prior to the first frame. A difference between the first unvoicing parameter and the smoothed unvoicing parameter for the first subframe is computed and a unvoiced/voiced classification of the first frame is determined using the computed difference as a decision parameter. Further processing, such as Bandwidth extension (BWE) is performed on based on the classification of the first frame.

Abstract: Methods, apparatus, and computer readable media related to soliciting feedback from a user regarding one or more content parameters of a suggestion or other content provided by the automated assistant. The user's feedback may be used to influence future suggestions and/or other content subsequently provided, by the automated assistant in future dialog sessions, to the user and/or to other users. In some implementations, content is provided to a user by an automated assistant in a dialog session between the user and the automated assistant—and the automated assistant provides a prompt that solicits user feedback related to the provided content in a future dialog session between the user and the automated assistant. In some of those implementations, the prompt is provided following input from the user and/or output from the automated assistant, in the future dialog session, that is unrelated to the content provided in the previous dialog session.

Abstract: An audio encoding apparatus and method that encodes hybrid contents including an object sound, a background sound, and metadata, and an audio decoding apparatus and method that decodes the encoded hybrid contents are provided. The audio encoding apparatus may include a mixing unit to generate an intermediate channel signal by mixing a background sound and an object sound, a matrix information encoding unit to encode matrix information used for the mixing, an audio encoding unit to encode the intermediate channel signal, and a metadata encoding unit to encode metadata including control information of the object sound.

Abstract: A system and method performs wakeword detection and automatic speech recognition using the same acoustic model. A mapping engine maps phones/senones output by the acoustic model to phones/senones corresponding to the wakeword. A hidden Markov model (HMM) may determine that the wakeword is present in audio data; the HMM may have multiple paths for multiple wakewords or may have multiple models. Once the wakeword is detected, ASR is performed using the acoustic model.

Abstract: This disclosure relates generally to speech signal processing, and more particularly to method and system for improving recognition of disordered speech. In one embodiment, a method and a system for identifying extent of deviation in speech utterances of a user from a normal level, caused due to such impairments are provided. The system further makes appropriate modifications to generate utterances pertaining to healthy speech. This data may be fed as input to the speech recognition systems, as those systems can interpret the corrected data.

Abstract: A data processing arrangement is configured to obtain plurality of documents including sentences, analyze sentences of plurality of documents to determine specific domain associated with each of plurality of documents, tokenize sentences in each of plurality of documents to obtain plurality of tokens for each of plurality of documents, determine token coordinates of each of plurality of tokens, and interpret contextual meaning of each of tokens of plurality of tokens for each of plurality of documents.

Abstract: A computing device displays a data visualization interface and receives user selection of a data source and a natural language command directed to the data source. The device forms a first intermediate expression according to a context-free grammar and a semantic model of data fields in the data source. In accordance with a determination that the first intermediate expression omits sufficient information for generating a data visualization, the device infers the omitted information associated with the data source using one or more inferencing rules based on syntactic and semantic constraints imposed by the context-free grammar. The device forms an updated intermediate expression, and translates the updated intermediate expression into database queries. It executes the database queries to retrieve data sets from the data source, then generates and displays a data visualization of the retrieved data sets.

Abstract: Implementations described herein relate to providing suggestions, via a display modality, for completing a spoken utterance for an automated assistant, in order to reduce a frequency and/or a length of time that the user will participate in a current and/or subsequent dialog session with the automated assistant. A user request can be compiled from content of an ongoing spoken utterance and content of any selected suggestion elements. When a currently compiled portion of the user request (from content of a selected suggestion(s) and an incomplete spoken utterance) is capable of being performed via the automated assistant, any actions corresponding to the currently compiled portion of the user request can be performed via the automated assistant. Furthermore, any further content resulting from performance of the actions, along with any discernible context, can be used for providing further suggestions.

Abstract: A method may include generating a source transaction description, encoding, by an encoder model of a machine translation model executing on a computer processor, the source transaction description to create a context vector, decoding, by a decoder model of the machine translation model, the context vector to predict a target entity description, generating a transaction including the target entity description, detecting an acceptance, by a user, of an action performed on the transaction, in response to detecting the acceptance, updating a translation accuracy metric for the target entity description, determining that the updated translation accuracy metric satisfies a translation accuracy criterion, and in response to determining that the updated translation accuracy metric satisfies the translation accuracy criterion, adding the target entity description to golden entity descriptions.

Abstract: A method and a system for dialogue enhancement of an audio signal, comprising receiving (step S1) the audio signal and a text content associated with dialogue occurring in the audio signal, generating (step S2) parameterized synthesized speech from the text content, and applying (step S3) dialogue enhancement to the audio signal based on the parameterized synthesized speech. With the invention text captions, subtitles, or other forms of text content included in an audio stream, can be used to significantly improve dialogue enhancement on the playback side.

Abstract: A method for voice recognition to perform an action. The method includes receiving a voice command, identifying a first action intended word from the voice command, assigning a confidence score to the first action intended word, altering the confidence score of the first action intended word in a temporal manner, based on confidence scores of second action intended words following the first action intended word in the voice command, identifying the action when the confidence scores of the first and second action intended words reach a pre-determined confidence score associated therewith, and performing the identified action. Disclosed also is a system for voice recognition to perform an action. The system includes one or more voice-controlled devices, and a voice managing server communicably coupled to the one or more voice-controlled devices. The voice managing server for voice recognition to perform an action using the aforementioned method.

Abstract: The present teaching relates to method, system, medium, and implementations for enabling communication with a user. Information representing surrounding of a user to be engaged in a new dialogue is received via the communication platform, wherein the information is acquired from a scene in which the user is present and captures characteristics of the user and the scene. Relevant features are extracted from the information. A state of the user is estimated based on the relevant features, and a dialogue context surrounding the scene is determined based on the relevant features. A topic for the new dialogue is determined based on the user, and a feedback is generated to initiate the new dialogue with the user based on the topic, the state of the user, and the dialogue context.

Abstract: A system and method for training a virtual agent to identify a user's intent from a conversation is disclosed. The system and method use an iterative process of clustering multiple conversations (converted into feature representations) used for training a machine learning model into labeled clusters having similar user intents. Clustering enables labeling a large number of training conversations efficiently. The labeled clusters may be used to train a virtual agent to classify the conversational intent of a conversation. Then, the machine learning model can classify future conversations based on similarity to labeled clusters. By knowing a human user's intent, a virtual agent can deliver what the user desires.