Abstract: Systems and methods for training a machine-learned model are provided. A method can include can include obtaining an unlabeled audio signal, sampling the unlabeled audio signal to select one or more sampled slices, inputting the one or more sampled slices into a machine-learned model, receiving, as an output of the machine-learned model, one or more determined characteristics associated with the audio signal, determining a loss function for the machine-learned model based at least in part on a difference between the one or more determined characteristics and one or more corresponding ground truth characteristics of the audio signal, and training the machine-learned model from end to end based at least in part on the loss function. The one or more determined characteristics can include one or more reconstructed portions of the audio signal temporally adjacent to the one or more sampled slices or an estimated distance between two sampled slices.

Abstract: An intention identification method includes generating a heterogeneous text network based on a language material sample; using a graph embedding algorithm to perform learning with respect to the heterogeneous text network and obtain a vector representation of the language material sample and a word, and determining keywords of the language material sample based on a similarity in terms of a vector between the language material sample and the word in the language material sample; training an intention identification model until a predetermined training termination condition is satisfied, by using the keywords of the language material samples, and obtaining the trained intention identification model; and receiving a language material query, and using the trained intention identification model to identify an intention of the language material query.

Abstract: A machine-learning method for training a decision-making model includes: obtaining a rationale vector group for a rationale included in a labeled natural language text file; assembling an effective vector group for the labeled natural language text file by connecting the rationale vector groups for the rationales using a specific order; and executing a supervised classification algorithm to train the decision-making model using the effective vector group and a target decision for the natural language text file. The decision-making model is trained to be configured to label an unlabeled natural language text file using one of a plurality of potential decisions that serves as a target decision.

Abstract: Methods, apparatus, systems, and computer-readable media are provided for using shortcut command phrases to operate an automated assistant. A user of the automated assistant can request that a shortcut command phrase be established for causing the automated assistant to perform a variety of different actions. In this way, the user does not necessarily have to provide an individual command for each action to be performed but, rather, can use a shortcut command phrase to cause the automated assistant to perform the actions. The shortcut command phrases can be used to control peripheral devices, IoT devices, applications, websites, and/or any other apparatuses or processes capable of being controlled through an automated assistant.

Abstract: A storage unit stores a target term, a substitute term, a substitute translated term, and a representative term. The substitute translated term is a translation of the substitute term and is expressed in a second language. The representative term indicates a type of the target term and is expressed in the second language. A communication unit acquires a provisional translation that is a translation of a processed sentence from a first external device that has a translation function. When the storage unit does not store a target translated term that is a translation of the target term, a controller replaces the substitute translated term contained in the provisional translation with the representative term to generate a second display-purpose translated sentence, and then causes a display unit to display the second display-purpose translated sentence.

Abstract: Training and/or using a multilingual classification neural network model to perform a natural language processing classification task, where the model reuses an encoder portion of a multilingual neural machine translation model. In a variety of implementations, a client device can generate a natural language data stream from a spoken input from a user. The natural language data stream can be applied as input to an encoder portion of the multilingual classification model. The output generated by the encoder portion can be applied as input to a classifier portion of the multilingual classification model. The classifier portion can generate a predicted classification label of the natural language data stream. In many implementations, an output can be generated based on the predicted classification label, and a client device can present the output.

Abstract: A gain control system for controlling gain applied to an audio signal includes a power estimator configured to estimate the power of a digital signal derived from the audio signal, a digital gain estimator configured to determine, in dependence on the estimated power, a digital gain which would modify the power of the digital signal so as to reach a target power level, and a gain controller configured to adjust an analogue gain applied to the audio signal in dependence on the determined digital gain.

Abstract: A method is provided for processing random interaction data. The method includes receiving random interaction data sent by a first terminal, and adding the random interaction data to a resource pool. The random interaction data carries user attribute information and inputted conversation information of a user of the first terminal. The method also includes, when no reply message for the random interaction data is received from one or more second terminals, activating a second interface and sending the random interaction data to an intelligent dialog system via the second interface, where the intelligent dialog system is configured to automatically generate reply information according to at least the inputted conversation information and the user attribute information. The method also includes receiving the reply information returned by the intelligent dialog system; and sending the reply information to the first terminal.

Abstract: Techniques for EMD-based signal de-noising are disclosed that use statistical characteristics of IMFs to identify information-carrying IMFs for the purposes of partially reconstructing the identified relevant IMFs into a de-noised signal. The present disclosure has identified that the statistical characteristics of IMFs with noise tend to follow a generalized Gaussian distribution (GGD) versus only a Gaussian or Laplace distribution. Accordingly, a framework for relevant IMF selection is disclosed that includes, in part, performing a null hypothesis test against a distribution of each IMF derived from the use of a generalized probability density function (PDF). IMFs that contribute more noise than signal may thus be identified through the null hypothesis test. Conversely, the aspects and embodiments disclosed herein enable the determination of which IMFs have a contribution of more signal than noise.

Abstract: A method for separating an audio input signal into a dry signal component and an ambient signal component is provided. The method includes generating a transferred input signal including transferring the audio input signal into frequency space and applying a smoothing filter to the transferred input signal to generate an estimated ambient signal component. The method includes determining the dry signal component based on the estimated ambient signal component and determining the ambient signal component based on the determined dry signal component and the audio input signal.

Abstract: An intelligent assistant records speech spoken by a first user and determines a self-selection score for the first user. The intelligent assistant sends the self-selection score to another intelligent assistant, and receives a remote-selection score for the first user from the other intelligent assistant. The intelligent assistant compares the self-selection score to the remote-selection score. If the self-selection score is greater than the remote-selection score, the intelligent assistant responds to the first user and blocks subsequent responses to all other users until a disengagement metric of the first user exceeds a blocking threshold. If the self-selection score is less than the remote-selection score, the intelligent assistant does not respond to the first user.

Abstract: A technique for estimating phonemes for a word written in a different language is disclosed. A sequence of graphemes of a given word in a source language is received. The sequence of the graphemes in the source language is converted into a sequence of phonemes in the source language. One or more sequences of phonemes in a target language are generated from the sequence of the phonemes in the source language by using a neural network model. One sequence of phonemes in the target language is determined for the given word. Also, technique for estimating graphemes of a word from phonemes in a different language is disclosed.

Abstract: A speech recognition method includes receiving a sentence generated through speech recognition, calculating a degree of suitability for each word in the sentence based on a relationship of each word with other words in the sentence, detecting a target word to be corrected among the words in the sentence based on the degree of suitability for each word, and replacing the target word with any one of candidate words corresponding to the target word.

Abstract: A control system is operable within a host vehicle to control the operation of signaling apparatus indicative of a driver intent to present a tissue box. The current invention permits the activation of an electro-mechanical system located proximally to said driver to permit access to a tissue. The control system includes a voice recognition circuit for activating a panel signal devices within the vehicle. In some embodiments, a wireless link facilitates aftermarket applications while in other embodiments original equipment manufacturer is accommodated. In some embodiments the tissue panel is opened manually.

Abstract: According to an embodiment, a voice recognition device includes one or more processors. The one or more processors are configured to: recognize a voice signal representing a voice uttered by an object speaker, to generate text and meta information representing information that is not included in the text and included in the voice signal; generate an object presentation vector including a plurality of parameters representing a feature of a presentation uttered by the object speaker; calculate a similarity between the object presentation vector and a reference presentation vector including a plurality of parameters representing a feature of a presentation uttered by a reference speaker; and output the text. The one or more processors are further configured to determine whether to output the meta information based on the similarity, and upon determining to output the meta information, add the meta information to the text and output the meta information.

Abstract: Methods and systems for automatic speech recognition and methods and systems for training acoustic language models are disclosed. In accordance with one automatic speech recognition method, an acoustic input data set is analyzed to identify portions of the input data set that conform to a general language and to identify portions of the input data set that conform to at least one dialect of the general language. In addition, a general language model and at least one dialect language model is applied to the input data set to perform speech recognition by dynamically selecting between the models in accordance with each of the identified portions. Further, speech recognition results obtained in accordance with the application of the models is output.

Abstract: Aspects of the present disclosure relate to voice command filtering. One or more directions of background noise for a location of a voice command device are determined. The one or more directions of background noise are stored as one or more blocked directions. A voice input is received at the location of the voice command device. A direction the voice input is being received from is determined and compared to the one or more blocked directions. The voice input is ignored in response to the direction of the voice input being received from corresponding to a direction of the one or more blocked directions, unless the received voice input is in a recognized voice.

Abstract: Contextual spelling methods and systems are provided that utilize natural language processing and n-gram frequencies to group documents into logical groups and to provide spelling correction suggestions. For example, a contextual spelling correction system may receive a set of documents, group the documents into separate logical groups, generate dictionaries associated with the logical groups, receive a user input, determine scores for potential spelling correction suggestions regarding the user input, and provide spelling correction suggestions based at least partly on the dictionaries associated with the logical groups.

Abstract: A system and method capable of distinguishing sources in a multiple source environment is provided. The system receives an audio signal comprising an audio tag, a desired audio signal and an undesired audio signal. Based on the audio tag, the system eliminates the undesired audio signal and identifies an intended command in the desired audio signal. The system generates a command for an external device based on the intended command.

Abstract: A technique for estimating phonemes for a word written in a different language is disclosed. A sequence of graphemes of a given word in a source language is received. The sequence of the graphemes in the source language is converted into a sequence of phonemes in the source language. One or more sequences of phonemes in a target language are generated from the sequence of the phonemes in the source language by using a neural network model. One sequence of phonemes in the target language is determined for the given word. Also, technique for estimating graphemes of a word from phonemes in a different language is disclosed.