Abstract: Methods and systems for natural language processing include generating an encoder that includes a global part and a local part, where the global part encodes multi-hop relations between words in an input and where the local part encodes one-hop relations between words in the input. The encoder is trained to form a graph that represents tokens of an input text as nodes and that represents relations between the tokens as edges between the nodes.

Abstract: Systems and methods are described for generating indications of real-time communication sessions. An ongoing communication session is monitored to identify a most recent subset of communications, the most recent subset of communications being defined by a sliding window. The most recent subset of communications is analysed to identify one or more relevant words, based on at least a user-specific relevancy criterion, the user-specific relevancy criterion being relevant to a user-specific topic associated with a given user profile. Responsive to identifying the one or more relevant words, an indication of the ongoing communication session is provided to a user device associated with the given user profile.

Abstract: A method of detecting establishment of a voice communication between a first voice communication equipment and a second voice communication equipment and automating requests for content. The method includes analyzing the voice communication to identify a request for content, analyzing the voice communication to identify an affirmative response to the request for content, and correlating the request for content with a first user account and correlating the affirmative response with a second user account. In response to identifying the affirmative response and based upon at least one of the first user account or the second user account, identifying from a data storage, the requested content and causing the transmission of the requested content.

Abstract: Data processing method, apparatus, electronic device, and computer storage medium are provided. The data processing method is used to generate a description information file related to a target object, and includes: obtaining a description framework and multiple types of multiple materials related to the target object, the description framework including attribute selection information corresponding to the target object; performing at least one type of processing on each material to obtain attribute information of the respective material, and the attribute information including an attribute level and an attribute content; selecting target materials whose attribute content and attribute level match the attribute selection information of the description framework; and generating a description information file according to the description framework and the target materials. The data processing method may automatically generate description information files.

Abstract: Devices and techniques are generally described for sending a first instruction for a device to output first content while the speech-processing device is in an ambient state during a first time period. First feedback data is received indicating that a first action associated with the first content was requested at a first time. A determination is made that the first time is during the first time period. Timing data related to a current time of the device is determined. Second content is determined based at least in part on the first action being requested during the first time period and the timing data. A second instruction is sent effective to cause the device to output second content while in the ambient state during a second time period.

Abstract: A method implemented via execution of computing instructions configured to run at one or more processors and stored at one or more non-transitory computer-readable media. The method can include receiving, via a computer network, a user interaction signal from a user device for a user. The method further can include after receiving the user interaction signal, determining, in real-time via a machine learning model, a plurality of user intent labels based at least in part on transaction data, interaction data, and incident data for the user. The machine learning model can include pre-trained based on historical input data and historical output data associated with multiple users comprising the user. The historical input data can comprise historical feature embedding vectors for historical transaction data, historical interaction data, and historical incident data associated with the multiple users.

Abstract: Techniques for providing device functionalities using device components are described. A system receives a system-generated directive from a skill system and determines a workflow to execute. The system implements a response orchestrator that operates based on the workflow that includes interception points where cross-cutting functionalities can be invoked as pluggable components. The interception points occur pre-system-generated directive, pre-device-facing directive, post-device-facing directive generation, post-device-facing directive dispatch, and the like. The system supports asynchronous execution, conditional execution, and sequential execution of components. Data determined by the cross functionality components can be used by other components for processing.

Abstract: Processor(s) of a client device can: receive sensor data that captures environmental attributes of an environment of the client device; process the sensor data using a machine learning model to generate a predicted output that dictates whether one or more currently dormant automated assistant functions are activated; making a decision as to whether to trigger the one or more currently dormant automated assistant functions; subsequent to making the decision, determining that the decision was incorrect; and in response to determining that the determination was incorrect, generating a gradient based on comparing the predicted output to ground truth output. In some implementations, the generated gradient is used, by processor(s) of the client device, to update weights of the on-device speech recognition model. In some implementations, the generated gradient is additionally or alternatively transmitted to a remote system for use in remote updating of global weights of a global speech recognition model.

Abstract: The present disclosure relates to a relay device of a wireless network, said wireless network comprising a plurality of network nodes mutually connected by wireless links. The relay device is configured to receive a voice command by a microphone of a source node, determine a recipient voice assistant for processing the voice command, and transmit, towards the recipient voice assistant, an output signal comprising the voice command. The present disclosure relates also to a voice assistant and to a wireless network, and to methods for processing voice commands by a relay device and a voice assistant.

Abstract: An embodiment dialogue system includes a speech recognizer configured to convert an utterance of a user into an utterance text, a natural language understanding module configured to identify an intention of the user based on the utterance text, and a controller configured to generate a first control signal for performing control corresponding to the intention of the user, identify whether an additional control item related to the control corresponding to the intention of the user exists, and in response to the additional control item existing, generate a second control signal for displaying information about the additional control item on a display.

Abstract: An audio stream received from an audio transceiver. The audio stream is in an environment that includes audio of a first user. An acoustic communication of the first user is detected from the audio stream. An audio intent trigger of the first user is identified from the audio stream and based on the acoustic communication. An assistance action for the first user is initiated in response to the audio intent trigger and by a voice-based interface.

Abstract: A system for performing automated speech recognition (ASR) on audio data includes a queue manager to receive a request to perform ASR on audio data, add the request to a queue of incoming requests, and determine a queue depth representing a number of requests in the queue at a given time. The system also includes a load supervisor to receive the request and the queue depth from the queue manager and assign a service level for the request based on the queue depth. In addition, the system includes a speech-to-text converter to receive the assigned service level for the request from the load supervisor, select an ASR model for the request based on the received service level, receive the audio data associated with the request, and perform ASR on the audio data using the selected ASR model.

Abstract: A rule integration device includes determination means for converting, to natural language sentences, each of multiple management rules used by a management device for managing a management target, and determining whether or not the multiple management rules are combinable based on grammar relating to the converted natural language sentences; and combination means for generating a post-combination rule by combining the multiple management rules that have been determined to be combinable by the determination means.

Abstract: Computer-based natural language understanding of input and output for a computer interlocutor is improved using a method of classifying conversation segments from transcribed conversations. The improvement includes one or more methods of splitting transcribed conversations into groups related to a conversation ontology using metadata; identifying dominant paths of conversational behavior by counting the frequency of occurrences of the behavior for a given path; creating a conversation model comprising conversation behaviors, metadata, and dominant paths; and using the conversation model to assign a probability score for a matched input to the computer interlocutor or a generated output from the computer interlocutor.

Abstract: Disclosed are a method and an apparatus for detecting a voice end point by using acoustic and language modeling information to accomplish strong voice recognition. A voice end point detection method according to an embodiment may comprise the steps of: inputting an acoustic feature vector sequence extracted from a microphone input signal into an acoustic embedding extraction unit, a phonemic embedding extraction unit, and a decoder embedding extraction unit, which are based on a recurrent neural network (RNN); combining acoustic embedding, phonemic embedding, and decoder embedding to configure a feature vector by the acoustic embedding extraction unit, the phonemic embedding extraction unit, and the decoder embedding extraction unit; and inputting the combined feature vector into a deep neural network (DNN)-based classifier to detect a voice end point.

Abstract: In various embodiments, a computer-implemented language identification and communication system is provided. The system includes an application engine configured for processing data associated with multiple access devices of a population of users who are users seeking human language translation services and users providing human language translation services. A geolocation module is provided for locating a position of various users, such as the different locations of human language translators. The application engine is further programmed for receiving translator selections from user access devices and establishing communication connections between or among different user access devices.

Abstract: Disclosed are a voice recognition system and a display device using the same. The disclosed voice recognition system includes a plate structure, a vibration sensor, and a voice recognition device. The plate structure vibrates based on propagation of a voice wave generated from a user, and the vibration sensor is provided in contact with the plate structure to detect the vibration of the plate structure. The voice recognition device recognizes voice of the user by receiving a signal output from the vibration sensor.

Abstract: A method for automatically rectifying in real-time anomalies in natural language processing systems. The method can include determining an output corresponding to a user request from a user device for a user based on a new request template or machine learning. The method further can include retrieving one or more entity rules corresponding to entity data of the user request. The method also can include overwriting entity information of the entity data corresponding to the one or more entity rules. Additionally, the method can include outputting the output. Furthermore, the method can include transmitting, to the user device, a response to the user. Other embodiments are disclosed.

Abstract: A device performs a method for using image data to aid voice recognition. The method includes the device capturing image data of a vicinity of the device and adjusting, based on the image data, a set of parameters for voice recognition performed by the device. The set of parameters for the device performing voice recognition include, but are not limited to: a trigger threshold of a trigger for voice recognition; a set of beamforming parameters; a database for voice recognition; and/or an algorithm for voice recognition. The algorithm may include using noise suppression or using acoustic beamforming.

Abstract: A system and method are disclosed for ignoring a wakeword received at a speech-enabled listening device when it is determined the wakeword is reproduced audio from an audio-playing device. Determination can be by detecting audio distortions, by an ignore flag sent locally between an audio-playing device and speech-enabled device, by and ignore flag sent from a server, by comparison of received audio played audio to a wakeword within an audio-playing device or a speech-enabled device, and other means.