Abstract: Systems and methods for providing audio data, from an initially invoked automated assistant to a subsequently invoked automated assistant. An initially invoked automated assistant may be invoked by a user utterance, followed by audio data that includes a query. The query is provided to a secondary automated assistant for processing. Subsequently, the user can submit a query that is related to the first query. In response, the initially invoked automated assistant provides the query to the secondary automated assistant in lieu of providing the query to other secondary automated assistants based on similarity between the first query and the subsequent query.

Abstract: Disclosed is an electronic device including processor and memory operatively connected to the processor and storing language model. The electronic device may enter data into the language model, generate an embedding vector in the input embedding layer, add position information to the embedding vector in the positional encoding layer, branch the embedding vector based on domain information, normalize the branched embedding vectors, enter the normalized embedding vectors into the multi-head attention layer, enter output data of the multi-head attention layer into the first layer, normalize pieces of output data of the first layer, enter the normalized pieces of output data of the first layer into the feed-forward layer, enter output data of the feed-forward layer into the second layer and normalize pieces of output data of the second layer, and enter the normalized pieces of output data of the second layer into the linearization layer and the softmax layer to obtain result data.

Abstract: Disclosed herein are system, apparatus, article of manufacture, method and/or computer program product embodiments, and/or combinations and sub-combinations thereof, for providing voice control using multiple digital assistants. In some embodiments, a voice platform operates to receive a voice input from a user. The voice platform selects a digital assistant from a plurality of digital assistants based on a trigger word. The voice platform then generates an intent from the voice input using the selected digital assistant. The voice platform then transmits the intent to a media device for processing.

Abstract: Embodiments described herein provide for audio processing operations that evaluate characteristics of audio signals that are independent of the speaker's voice. A neural network architecture trains and applies discriminatory neural networks tasked with modeling and classifying speaker-independent characteristics. The task-specific models generate or extract feature vectors from input audio data based on the trained embedding extraction models. The embeddings from the task-specific models are concatenated to form a deep-phoneprint vector for the input audio signal. The DP vector is a low dimensional representation of the each of the speaker-independent characteristics of the audio signal and applied in various downstream operations.

Abstract: The present document relates to audio coding systems which make use of a harmonic transposition method for high frequency reconstruction (HFR), and to digital effect processors, e.g. so-called exciters, where generation of harmonic distortion adds brightness to the processed signal. In particular, a system configured to generate a high frequency component of a signal from a low frequency component of the signal is described. The system may comprise an analysis filter bank (501) configured to provide a set of analysis subband signals from the low frequency component of the signal; wherein the set of analysis subband signals comprises at least two analysis subband signals; wherein the analysis filter bank (501) has a frequency resolution of ?f.

Abstract: Systems and methods for utilizing microphone array information for acoustic modeling are disclosed. Audio data may be received from a device having a microphone array configuration. Microphone configuration data may also be received that indicates the configuration of the microphone array. The microphone configuration data may be utilized as an input vector to an acoustic model, along with the audio data, to generate phoneme data. Additionally, the microphone configuration data may be utilized to train and/or generate acoustic models, select an acoustic model to perform speech recognition with, and/or to improve trigger sound detection.

Abstract: Techniques are disclosed relating to implementing a machine learning approach to cross-language translation and search. In certain embodiments, a method may include receiving a plurality of characters of a first language that are unsegmented and grouping the plurality of character into multiple groups. The method also includes determining a set of word tokens based on one or more transliterations of the multiple groups and one or more translations of the multiple groups to a second language. Further, the method includes generating one or more word token solution sets by querying an index file using the one or more word tokens. The method also includes determining whether the index file references an entity name corresponding to the plurality of characters of the first language based on comparing the one or more token solution sets with the index file.

Abstract: A system and method for dictation using a peripheral device includes a voice recognition mouse. The voice recognition mouse includes a microphone, a first button, a processor coupled to the microphone and the first button, and a memory coupled to the processor. The memory stores instructions that, when executed by the processor, cause the processor to detect actuation of the first button and in response to detecting actuation of the first button, invoke the microphone for capturing audio speech from a user. The captured audio speech is streamed to a first module. The first module is configured to invoke a second module for converting the captured audio speech into text and forward the text to the first module for providing to an application expecting the text, the application being configured to display the text on a display device.

Abstract: Coordinated operation of a voice-controlled device and an accessory device in an environment is described. A remote system processes audio data it receives from the voice-controlled device in the environment to identify a first intent associated with a first domain, a second intent associated with a second domain, and a named entity associated with the audio data. The remote system sends, to the voice-controlled device, first information for accessing main content associated with the named entity, and a first instruction corresponding to the first intent. The remote system also sends, to the accessory device, second information for accessing control information or supplemental content associated with the main content, and a second instruction corresponding to the second intent. The first and second instructions, when processed by the devices in the environment, cause coordinated operation of the voice-controlled device and the accessory device.

Abstract: At least selectively utilizing a large language model (LLM) in generating a natural language (NL) based summary to be rendered in response to a query. In some implementations, in generating the NL based summary additional content is processed using the LLM. The additional content is in addition to query content of the query itself and, in generating the NL based summary, can be processed using the LLM and along with the query content—or even independent of the query content. Processing the additional content can, for example, mitigate occurrences of the NL based summary including inaccuracies and/or can mitigate occurrences of the NL based summary being over-specified and/or under-specified.

Abstract: A system and method for processing and actionizing structured and unstructured experience data is disclosed herein. In some embodiments, a system may include a natural language processing (NLP) engine configured to transform a data set into a plurality of concepts within a plurality of distinct contexts, and a data mining engine configured to process the relationships of the concepts and to identify associations and correlations in the data set. In some embodiments, the method may include the steps of receiving a data set, scanning the data set with an NLP engine to identify a plurality of concepts within a plurality of distinct contexts, and identifying patterns in the relationships between the plurality of concepts.

Abstract: A method for recognizing a Chinese-English mixed speech, includes: determining pronunciation information and scores of a language model, of speech information, in response to receiving the speech information; determining whether an English word exists in content of the speech information based on the pronunciation information; determining a Chinese word corresponding to the English word based on a preset Chinese-English mapping table in response to the English word existing in the content of the speech information, in which the Chinese-English mapping table includes a mapping relationship of at least one pair of English word and Chinese word; determining a score of the Chinese word corresponding to the English word; replacing a score of the English word in the scores of the language model with the score of the Chinese word; and obtaining a speech recognition result for the speech information based on the replaced scores of the language model.

Abstract: For a seamless and robust artificial intelligence-based assistant experience, an intent-based query and response router has been designed to operate as an intelligent layer between a user and multiple backend services that may respond to one or more queries over the course of a conversation with the user. The query router interacts with an intent classification service to obtain an intent classification for a prompt that is based on a user query. The query router uses the intent classification, which is used as an identifier of a backend service, to route the user query to an appropriate one (or more) of the backend services. When a response is detected, the query router determines a corresponding conversation and provides the response for the conversation.

Abstract: Systems and processes for operating an intelligent automated assistant are provided. In one example process, a first input including activation of an affordance is received. A domain associated with the affordance is determined. A second input including user speech is received, where a user intent is determined based on the domain and the user speech. A determination is made whether the user intent includes a command associated with the affordance. In accordance with a determination that the user intent includes a command associated with the affordance, a task in furtherance of the command is performed.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for hotword detection on multiple devices are disclosed, In one aspect, a method includes the actions of receiving audio data that corresponds to an utterance. The actions further include determining that the utterance likely includes a particular, predefined hotword. The actions further include transmitting (i) data indicating that the computing device likely received the particular, predefined hotword, (ii) data identifying the computing device, and (iii) data identifying a group of nearby computing devices that includes the computing device. The actions further include receiving an instruction to commence speech recognition processing on the audio data. The actions further include in response to receiving the instruction to commence speech recognition processing on the audio data, processing at least a portion of the audio data using an automated speech recognizer on the computing device.

Abstract: An off-policy reinforcement learning actor-critic neural network system configured to select actions from a continuous action space to be performed by an agent interacting with an environment to perform a task. An observation defines environment state data and reward data. The system has an actor neural network which learns a policy function mapping the state data to action data. A critic neural network learns an action-value (Q) function. A replay buffer stores tuples of the state data, the action data, the reward data and new state data. The replay buffer also includes demonstration transition data comprising a set of the tuples from a demonstration of the task within the environment. The neural network system is configured to train the actor neural network and the critic neural network off-policy using stored tuples from the replay buffer comprising tuples both from operation of the system and from the demonstration transition data.

Abstract: The present invention relates to a context-based QA generation architecture, and an object of the present invention is to generate diverse QA pairs from a single context. To achieve the object, the present invention includes a latent variable generating network including at least one encoder and an artificial neural network (Multi-Layer Perceptron: MLP) and configured to train the artificial neural network using a first context, a first question, and a first answer, and generate a second question latent variable and a second answer latent variable by applying the trained artificial neural network to a second context, an answer generating network configured to generate a second answer by decoding the second answer latent variable, and a question generating network configured to generate a second question based on a second context and the second answer.

Abstract: At least selectively utilizing a large language model (LLM) in generating a natural language (NL) based summary to be rendered in response to a query. In some implementations, in generating the NL based summary additional content is processed using the LLM. The additional content is in addition to query content of the query itself and, in generating the NL based summary, can be processed using the LLM and along with the query content—or even independent of the query content. Processing the additional content can, for example, mitigate occurrences of the NL based summary including inaccuracies and/or can mitigate occurrences of the NL based summary being over-specified and/or under-specified.

Abstract: A method for generating a prosodic representation includes receiving a text utterance having one or more words. Each word has at least one syllable having at least one phoneme. The method also includes generating, using a Bidirectional Encoder Representations from Transformers (BERT) model, a sequence of wordpiece embeddings and selecting an utterance embedding for the text utterance, the utterance embedding representing an intended prosody. Each wordpiece embedding is associated with one of the one or more words of the text utterance. For each syllable, using the selected utterance embedding and a prosody model that incorporates the BERT model, the method also includes generating a corresponding prosodic syllable embedding for the syllable based on the wordpiece embedding associated with the word that includes the syllable and predicting a duration of the syllable by encoding linguistic features of each phoneme of the syllable with the corresponding prosodic syllable embedding for the syllable.

Abstract: A decoder for generating an audio output signal having one or more audio output channels is provided, having a receiving interface for receiving an audio input signal having a plurality of audio object signals, for receiving loudness information on the audio object signals, and for receiving rendering information indicating whether one or more of the audio object signals shall be amplified or attenuated, further having a signal processor for generating the one or more audio output channels of the audio output signal, configured to determine a loudness compensation value depending on the loudness information and depending on the rendering information, and configured to generate the one or more audio output channels of the audio output signal from the audio input signal depending on the rendering information and depending on the loudness compensation value. One or more by-pass audio object signals are employed for generating the audio output signal. Moreover, an encoder is provided.