Abstract: Techniques for fine-tuning language models using target language vocabulary and parallel data for machine translation are disclosed. In an example method, a computing device receives one or more words in a source language. The computing device translates the one or more words to a target language using a fine-tuned language model. The language model is fine-tuned by accessing a foundation language model pre-trained using source language training data and having a source language vocabulary. A set of tokens is added to the model representing a target language vocabulary. The model is further trained using target language training data. The model is further trained using parallel training data including a number of parallel elements each parallel element including words in the source language and the corresponding words in the target language. The computing device outputs the translated one or more words in the target language.

Abstract: Methods and systems for using voice (e.g., voice commands) to control a plurality of network devices via a motion sensing control device are provided. A control device can detect movement (e.g., a gesture) associated with the control device. Based on the movement satisfying a predefined condition, the control device can initiate a either a direct or a proxy communication session with a remote computing device. The communication session can be established and maintained for a predefined period such that data associated with a command can be immediately transmitted to the remote computing device. Thus, when a voice command is received, data associated with the command can be transmitted over the already established communication session to the remote computing device. The remote computing device can provide a response to the control device and/or transmit a command code associated with the voice command to one or more devices intended to be controlled.

Abstract: A speech-processing system may provide access to multiple virtual assistants via a user device. A user may invoke a particular virtual assistant by speaking its wakeword. The virtual assistants may send notifications to the user via the user device, which may present an indication of the available notifications. The user may request delivery of the notifications using, for example, a voice user interface. When delivering notifications from multiple virtual assistants, the system may use distinct voice and/or visual characteristics to indicate a source of the notifications. For example, after delivering a first notification corresponding to a first virtual assistant (e.g., the one invoked by the user at the beginning of the interaction) the first virtual assistant may offer to deliver a second notification from a second virtual assistant. The second virtual assistant may deliver the second notification using voice/visual characteristics different from those used for the first notification.

Abstract: Systems and methods for audio processing include capturing first sound data via at least one microphone of a network microphone device (NMD) and determining, via a voice activity detection process, that the first sound data does not include voice activity. The first sound data is stored in a buffer, and the NMD forgoes spatial processing of the first sound data. The NMD can capture second sound data and determine, via the voice activity process, that the second sound data includes voice activity. The NMD spatially processes the second sound data to produce filtered sound data. The NMD detects a wake word based on data in the buffer. After detecting the wake word, the NMD may determine an action to be performed based on the data in the buffer.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for sanitizing artificial intelligence prompts. One of the methods includes receiving a message a) for an external system and b) that comprises two or more phrases; for at least one phrase from the two or more phrases: determining a context of the phrase in the message; determining, using the context, whether modification of the phrase will likely maintain an intent of the message; determining whether to permit unedited transmission of the message to the external system using a result of at least one of one or more determinations whether modification of the phrase will likely maintain the intent of the message; and performing one or more actions using a result of the determination whether to permit unedited transmission of the message to the external system.

Abstract: A method and apparatus for encoding/decoding audio signal are provided. The encoding method includes transforming an input audio signal in a time domain into an audio signal in a frequency domain, quantizing energy of a frequency band of the audio signal in the frequency domain, generating a normal signal by normalizing the audio signal in the frequency domain according to quantized energy, obtaining a feature vector including information on the energy of the frequency band based on the normal signal and the input audio signal, quantizing the feature vector, obtaining a scale factor used to scale the normal signal based on the quantized feature vector, quantizing an adjustment signal into which the normal signal has been scaled based on the scale factor, and outputting bitstreams based on the quantized energy, the quantized feature vector, and the quantized adjustment signal.

Abstract: A method performed by an encoder. The method comprises determining envelope representation residual coefficients as first compressed envelope representation coefficients subtracted from the input envelope representation coefficients. The method comprises transforming the envelope representation residual coefficients into a warped domain so as to obtain transformed envelope representation residual coefficients. The method comprises applying, at least one of a plurality of gain-shape coding schemes on the transformed envelope representation residual coefficients in order to achieve gain-shape coded envelope representation residual coefficients, where the plurality of gain-shape coding schemes have mutually different trade-offs in one or more of gain resolution and shape resolution for one or more of the transformed envelope representation residual coefficients.

Abstract: The input of a user is monitored and a location of the user and a language of the user are detected. The input is converted to a text string in the detected user language and the converted text string is parsed into parsed tokens. A command line and correlated parameters indicated by the input are recognized based on the parsed tokens. The recognized command line with the assigned parameters is executed.

Abstract: An electronic device and a method of controlling the electronic device are provided. The electronic device includes a microphone, a display, memory, and a processor configured to obtain, based on a user voice being received through the microphone while a first user interface (UI) screen is being displayed, text information corresponding to the user voice by inputting the user voice, obtain first information including information on a command and information on an execution target of the command included in the text information, obtain second information including information on functions corresponding to the plurality of objects and information on texts included in the plurality of objects, identify whether a target object corresponding to the user voice is present from among the plurality of objects, and control the display to display a second UI screen corresponding to the target object for performing an operation corresponding to the command.

Abstract: In some implementations, a method includes displaying, on a display, an environment that includes a representation of a virtual agent that is associated with a sensory characteristic. In some implementations, the method includes selecting, based on the sensory characteristic associated with the virtual agent, a subset of a plurality of sensors to provide sensor data for the virtual agent. In some implementations, the method includes providing the sensor data captured by the subset of the plurality of sensors to the virtual agent in order to reduce power consumption of the device. In some implementations, the method includes displaying a manipulation of the representation of the virtual agent based on an interpretation of the sensor data by the virtual agent.

Abstract: A system for an automated real-time transcription and editing of audio data using interim text, including a processor of an audio transcription server (ATS) node configured to host a machine learning (ML) module coupled to at least one audio source entity and connected to at least one user-entity node over a network and a memory on which are stored machine-readable instructions that when executed by the processor, cause the processor to: acquire audio data from the at least one audio source entity; parse out the audio data to derive features for beam forming and features for speaker diarization; generate a set of classifiers based on the features for beam forming and the features for speaker diarization; provide the set of classifiers to the ML module configured to generate a predictive model for producing at least one speaker identification parameter; identify the speaker based on the at least one speaker identification parameter; continuously transcribe the audio data to generate an interim text associated

Abstract: Systems and processes for a multi-modal digital assistant are provided.

Abstract: A method for identifying and executing a voice command in a continuous listening Internet of Things (IoT) environment, may include: receiving, by at least one IoT device, a voice input in the continuous listening IoT environment; detecting, by the at least one IoT device, an occurrence of at least one non-speech event in a vicinity of at least one other IoT device in the continuous listening IoT environment; determining, by the at least one IoT device, an ambient context associated with the at least one non-speech event; determining, by the at least one IoT device, a correlation between the ambient context and the at least one other IoT device based on an event location of the occurrence of the at least one non-speech event; and determining, by the at least one IoT device, presence of at least one voice command within the voice input based on the correlation.

Abstract: A speech synthesis system is described and may include at least one microphone; a speaker; a sensing system, and memory storing processor-executable instructions, which when executed by the processor, cause the processor to: detect speech-related signals emanating from the subject; generate a variable excitation signal; shape the generated variable excitation signal according to previously stored speech recordings; and cause, from the speaker and based on the shaped variable excitation signal, produced speech content that approximates the matched one or more voice characteristics in the previously stored speech recordings.

Abstract: The present disclosure relates to scalable systems and methods for detecting, labeling, and protecting sensitive data in natural language processing (NLP) environments. This includes NLP applications in artificial intelligence (AI) systems, such as language models (LMs) and generative AI (GenAI). More particularly, the present disclosure introduces a hierarchical, context-aware labeling mechanism that is optimized using an LM in conjunction with machine learning (ML) techniques to ensure the utility-preserving effective protection of sensitive data with, for example, minimal false positives and false negatives and/or optimal precision and recall (e.g., in terms of an F1 Score).

Abstract: In aspects of presenting relevant audio data, a mobile device implements an audio playback manager that monitors audio in an environment for a trigger word. The audio playback manager detects the trigger word via a microphone associated with a headset in communication with a mobile device. The audio playback manager determines whether a portion of the audio preceding the trigger word is relevant to a user of the mobile device and presents the portion of the audio that is relevant to the user.

Abstract: An embodiment computer-implemented method for predicting an intention of a user includes receiving from a vehicle first utterance data obtained by converting a voice command of the user into text, performing natural language understanding to attempt to decide the intention of the user from the first utterance data, predicting the intention of the user using stored pattern data in response to failing to decide the intention of the user, wherein the stored pattern data includes a plurality of patterns and confidence generated based on second utterance data received from each of a plurality of vehicles that are unable to decide the intention of the user and subsequent action data, generating a prompt suggesting an operation based on the predicted intention, and transmitting the prompt to the vehicle.

Abstract: In an embodiment, the disclosure relates to a device for assisting a respondent in a conversation. The device includes a microphone configured to detect a voice input, and a transmitter communicatively coupled to a server and configured to transmit the voice input to the server. The server is to generate vectors associated with the voice input, feed the vectors associated with the voice input to an Artificial Intelligence utilizing a trained Machine Learning (ML) model, and obtain, from the trained ML model, an output corresponding to the vectors. The device further includes a receiver communicatively coupled to the server, and configured to receive from the server, the output generated by the ML model. A speaker is communicatively coupled with the receiver and is configured to generate a voice-based response based on the output, for assisting the respondent in responding to the conversation.

Abstract: An approach is disclosed for enabling contextually relevant conversational interaction. Environment data is received by an AI System which detects a plurality of physical objects in a physical environment and forms a contextual understanding of the plurality of physical objects and the physical environment and identifies a user relevant to the contextual understanding. A most relevant contextual information to the user is predicted by the AI system and transformed into a textual form. A set of intents and objectives is predicted by the AI system for user-centered interaction. The AI system and the user interact iteratively through the user-centered interaction to determine an understanding of a most relevant intent and a most relevant objective which is validated by the AI system with the user until the user agrees. The validated most relevant intent and the most relevant objective is utilized to facilitate the user-centered and contextually relevant conversational interaction.

Abstract: A computer-executable Natural Language Understanding configuration generator is configured to generate a Natural Language Understanding configuration. The Natural Language Understanding configuration for a Natural Language Understanding component is based on engineering data related to an asset of an automation system.