Abstract: The present disclosure discloses a method for generating a styled sentence by a computer device. The method includes: obtaining a to-be-converted natural sentence, classifying, by inputting the natural sentence into a first encoding model, having a classification capability of classifying the natural sentence into a target content vector and a style vector of the natural sentence, the target content vector indicating a meaning of the natural sentence, and the style vector of the natural sentence indicating a language style of the natural sentence. The method also include determining, from at least one style vector according to a set target language style, a target style vector corresponding to the target language style; and inputting the target content vector and the target style vector into a first decoding model, and generating a styled sentence corresponding to the natural sentence.

Abstract: Systems and methods are described to address shortcomings in conventional systems by correcting an erroneous term in on-screen caption text for a media asset. In some aspects, the systems and methods identify the erroneous term in a text segment of the on-screen caption text, and identify one or more video frames of the media asset corresponding to the text segment. The systems and methods further identify a contextual term related to the erroneous term from the one or more video frames. By accessing a knowledge graph, the systems and methods identify a candidate correction based on the contextual term and a portion of the text segment. Lastly, the systems and methods replaces the erroneous term with the candidate correction.

Abstract: Medical records may be analyzed to identify important items in the text of the medical record. Actionable content may be identified and may be emphasized or extracted from the medical record. Actionable content may be categorized into one or more categories. Identification may include processing using trained models that use contextual information and position information to determine sentence labels.

Abstract: In one aspect, a playback device includes a voice assistant service (VAS) wake-word engine and a command keyword engine. The playback device detects, via the command keyword engine, a first command keyword of in voice input of sound detected by one or more microphones of the playback device. The playback device determines an intent based on at least one keyword in the voice input via a local natural language unit (NLU). After detecting the first command keyword event and determining the intent, the playback device performs a first playback command corresponding to the first command keyword and according to the determined intent. When the playback device detects, via the wake-word engine, a wake-word in voice input, the playback device streams sound data corresponding to at least a portion of the voice input to one or more remote servers associated with the VAS.

Abstract: Disclosed herein are embodiments of systems and methods for facilitating interpretation services by continuously generating a real time speed of speech for a set of spoken words within an electronic communication session. A processor continuously generates the real time speed of speech by iteratively generating the real time speed of speech upon parsing each spoken word from audio input signals of the electronic communication session. Iteratively generating the real time speed of speech may be based on a frequency of a subset of the set of spoken words within a preceding time interval of each spoken word. The systems and method may further determine a threshold based on an attribute of the set of spoken words. The method may display on a graphical user interface a graphical indicator having a visual feature that changes when the continuously generated real time speed of speech achieves the threshold.

Abstract: A method for maintenance of a machine among a fleet of machines includes receiving a service request corresponding to the machine. The method also includes obtaining a service architecture corresponding to the fleet of machines. The service architecture includes a service dictionary and a plurality of classification schemes organized in a tree data structure. The method also includes processing the service request based on the service dictionary and a text parsing technique to generate a list of descriptive words. The method includes generating a recommendation based on the list of descriptive words and the service architecture. The recommendation includes at least one of an on-line repair activity, an on-site repair activity and a part replacement activity. The method also includes servicing the fault condition of the machine based on the recommendation.

Abstract: A device control apparatus includes a detector which detects sound; and a controller which controls a plurality of devices. The plurality of devices includes a first device which is not related to operation of a vehicle and a second device which is related to the operation of the vehicle. The controller is configured to: recognize voice of a user by using data of the sound detected by the detector; identify a type of an operation target device from a plurality of the devices and operation content of the operation target device, based on the voice recognized; notify a user of start of activating the first device by the operation content identified and activate the first device by the operation content identified when the operation target identified is the first device; and notify the user to request a response before the controller activates the second device by the operation content identified when the operation target identified is the second device.

Abstract: Processor(s) of a client device can: receive audio data that captures a spoken utterance of a user of the client device; process, using an on-device speech recognition model, the audio data to generate a predicted textual segment that is a prediction of the spoken utterance; cause at least part of the predicted textual segment to be rendered (e.g., visually and/or audibly); receive further user interface input that is a correction of the predicted textual segment to an alternate textual segment; and generate a gradient based on comparing at least part of the predicted output to ground truth output that corresponds to the alternate textual segment. The gradient is used, by processor(s) of the client device, to update weights of the on-device speech recognition model and/or is transmitted to a remote system for use in remote updating of global weights of a global speech recognition model.

Abstract: An electronic apparatus is provided. The electronic apparatus includes a communicator, a memory, and a processor connected to the communicator and the memory and configured to control the electronic apparatus. The processor is configured to, by executing at least one command stored in the memory, based on a user input for executing an assistant service being received, transmit information on a user voice acquired by the electronic apparatus to a plurality of servers providing different assistant services through the communicator, and based on a plurality of response information being received from the plurality of servers, provide a response on the user voice based on at least one of the plurality of response information. The plurality of servers provide the assistant service using an artificial intelligence agent.

Abstract: Neural networks with similar architectures may be compressed using shared compression profiles. A request to compress a trained neural network may be received and an architecture of the neural network identified. The identified architecture may be compared with the different network architectures mapped to compression profiles to select a compression profile for the neural network. The compression profile may be applied to remove features of the neural network to generate a compressed version of the neural network.

Abstract: A noise cancellation method including generating a first voice signal by canceling a first portion of noise included in an input voice signal using a first network, the first network being a trained u-net structure, and the first portion of the noise being in a time domain, applying a first window to the first voice signal, performing a fast Fourier transform on the first windowed voice signal to acquire a magnitude signal and a phase signal, acquiring a mask using a second network based on the magnitude signal, the second network being another trained u-net structure, applying the mask to the magnitude signal, generating a second voice signal by canceling a second portion of the noise by performing an inverse fast Fourier transform on the first windowed voice signal based on the masked magnitude signal and the phase signal, and applying a second window to the second voice signal.

Abstract: A method includes receiving a set of text documents. The method also includes generating a summary of the set of text documents by a set of large language machine learning models. The method further includes generating a set of keywords from the summary by the set of large language machine learning models. The method additionally includes generating an image prompt from the set of keywords by the set of large language machine learning models. The method also includes generating a set of images from the image prompt by a text-to-image machine learning model. The method further includes generating a video clip from the set of images. The method additionally includes presenting the video clip.

Abstract: Systems and methods for providing notifications without breaking media immersion. A notification delivery application receives notification data while a media device provides a media asset. In response to receiving the notification data while the media device provides the media asset, the notification delivery application generates a voice model based on a voice detected in the media asset. The notification delivery application converts the notification data to synthesized speech using the voice model and generates, by the media device, the synthesized speech for output at an appropriate point in the media asset based on contextual features of the media asset.

Abstract: A system includes one or more field terminals, which can accept information inputs in voice and, optionally, other formats. The field terminals are equipped to translate voice inputs and to populate files according to specified requirements. Artificial intelligence algorithms augment speech to text translation thereby reducing translation errors and decreasing computing time. The field terminal completes the translation, populates the required document according to specified protocols and optionally attaches or associates information such as photos, location, video, maps or any other information that may be useful for the intended resource recipient(s). The field terminal comprises a transceiver, and transmits the file to the intended recipients.

Abstract: A speech-processing system may provide access to one or more virtual assistants via an audio-controlled device. A user may leverage a first virtual assistant to translate a natural language command from a first language into a second language, which the device can send to a second virtual assistant for processing. The device may receive a command from a user and send input data representing the command to a first speech-processing system representing the first virtual assistant. The device may receive a response in the form of a first natural language output from the first speech-processing system along with an indication that the first natural language output should be directed to a second speech-processing system representing the second virtual assistant. For example, the command may be in the first language, and the first natural language output may be in the second language, which is understandable by the second speech-processing system.

Abstract: Various embodiments contemplate systems and methods for performing automatic speech recognition (ASR) and natural language understanding (NLU) that enable high accuracy recognition and understanding of freely spoken utterances which may contain proper names and similar entities. The proper name entities may contain or be comprised wholly of words that are not present in the vocabularies of these systems as normally constituted. Recognition of the other words in the utterances in question, e.g. words that are not part of the proper name entities, may occur at regular, high recognition accuracy. Various embodiments provide as output not only accurately transcribed running text of the complete utterance, but also a symbolic representation of the meaning of the input, including appropriate symbolic representations of proper name entities, adequate to allow a computer system to respond appropriately to the spoken request without further analysis of the user's input.

Abstract: A method for processing sound that includes, generating one or more noise component estimates relating to an electrical representation of the sound and generating an associated confidence measure for the one or more noise component estimates. The method further comprises processing, based on the confidence measure, the sound.

Abstract: A voice-activated camera system for a computing device. The voice-activated camera system includes a processor, a camera module, a speech recognition module and a microphone for accepting user voice input. The voice-activated camera system includes authorized for only a specific user's voice, so that a camera function may be performed when the authorized user speaks the keyword, but the camera function is not performed when an unauthorized user speaks the keyword.

Abstract: Systems and methods of the present technical solution enable a multi-modal interface for voice-based devices, such as digital assistants. The solution can enable a user to interact with video and other content through a touch interface and through voice commands. In addition to inputs such as stop and play, the present solution can also automatically generate annotations for displayed video files. From the annotations, the solution can identify one or more break points that are associated with different scenes, video portions, or how-to steps in the video. The digital assistant can receive input audio signal and parse the input audio signal to identify semantic entities within the input audio signal. The digital assistant can map the identified semantic entities to the annotations to select a portion of the video that corresponds to the users request in the input audio signal.

Abstract: This application relates to microphone authentication apparatus for verifying whether or not an audio signal originated at a microphone. The microphone authentication apparatus has a comparison block configured to receive a first signal indicative of one or more spectral parameters of at least part of an audio signal to be verified, and compare the one or more spectral parameters to one or more predetermined characteristic microphone parameters relating to a characteristic resonance associated with an acoustic port of a microphone. The first signal may be an audio signal and the microphone authentication apparatus may have a feature extract module for determining the spectral parameters. Based on the comparison determination block may whether the audio signal originated from a microphone and may send a verification signal to a voice biometric module.