Abstract: Embodiments are directed to organizing conversations. Words may be provided from a conversation stream. Each word may be mapped to a graph model based on characteristics of each word. The graph model may be partitioned based on one or more attributes of a nodes and edges included in the graph model such that nodes associated with relationship strength that exceeds a threshold value may be grouped together. Sentence models may be generated based on sentences included in the conversation stream. Combined models may be generated based on the sentence models and the graph such that each sentence model may be associated with one or more partitions of the graph model. A conversation digest may be generated based on the combined model such that the conversation digest identifies one or more dominant portions of the conversation that include key subject matter.

Abstract: Disclosed is a method and apparatus with recognition for speaker recognition. The method includes determining whether an input feature vector corresponding to a voice signal of a speaker meets a candidate similarity criterion with at least one registered data included in a registration database, selectively, based on a result of the determining of whether the input feature vector meets the candidate similarity criterion, constructing a candidate list based on the input feature vector, determining whether a candidate input feature vector, among one or more candidate input feature vectors constructed in the candidate list in the selective constructing of the candidate list, meets a registration update similarity criterion with the at least one registered data, and selectively, based on a result of the determination of whether the candidate input feature vector meets the registration update similarity criterion, updating the registration database based on the candidate input feature vector.

Abstract: Systems and processes for operating a digital assistant are provided. In an example process, low-latency operation of a digital assistant is provided. In this example, natural language processing, task flow processing, dialogue flow processing, speech synthesis, or any combination thereof can be at least partially performed while awaiting detection of a speech end-point condition. Upon detection of a speech end-point condition, results obtained from performing the operations can be presented to the user. In another example, robust operation of a digital assistant is provided. In this example, task flow processing by the digital assistant can include selecting a candidate task flow from a plurality of candidate task flows based on determined task flow scores. The task flow scores can be based on speech recognition confidence scores, intent confidence scores, flow parameter scores, or any combination thereof. The selected candidate task flow is executed and corresponding results presented to the user.

Abstract: Disclosed herein are example techniques to support multiple voice assistant services. An example implementation may involve a playback device capturing audio from the one or more microphones into one or more buffers as a sound data stream monitoring the sound data stream for a wake word associated with a specific voice assistant service and monitoring the sound data stream for a wake word associated with the media playback system. The playback device generates a second wake-word event corresponding to a voice input when sound data matching the wake word associated with the media playback system in a portion of the sound data stream is detected. The playback device determines that the voice input includes sound data matching one or more playback commands and sends sound data representing the voice input to a voice assistant associated with the media playback system for processing of the second voice input.

Abstract: A method and a device for generating a modified statement are provided. An embodiment thereof includes obtaining a target statement, and separating words in the target statement to obtain a word collection to be modified; determining whether a target storage portion comprises a predetermined word collection to be confirmed corresponding to a word to be modified in the word collection to be modified; in response to determine to include the predetermined word collection, based on the error probabilities and the usage probabilities of the words to be confirmed, determining modification probabilities of the words to be confirmed; according to a numerical sequence of the modification probabilities, selecting predetermined numbered words to be confirmed from the word collection to be confirmed as a word sub-collection to be confirmed; and based on the word sub-collection to be confirmed, generating the modified statement. The embodiment improves the accuracy to modify the statement.

Abstract: Systems and processes for operating an intelligent automated assistant are provided. In accordance with one example, a method includes, at an electronic device with one or more processors, memory, and a plurality of microphones, sampling, at each of the plurality of microphones of the electronic device, an audio signal to obtain a plurality of audio signals; processing the plurality of audio signals to obtain a plurality of audio streams; and determining, based on the plurality of audio streams, whether any of the plurality of audio signals corresponds to a spoken trigger. The method further includes, in accordance with a determination that the plurality of audio signals corresponds to the spoken trigger, initiating a session of the digital assistant; and in accordance with a determination that the plurality of audio signals does not correspond to the spoken trigger, foregoing initiating a session of the digital assistant.

Abstract: Some implementations relate to performing speech biasing, NLU biasing, and/or other biasing based on historical assistant interaction(s). It can be determined, for one or more given historical interactions of a given user, whether to affect future biasing for (1) the given user account, (2) additional user account(s), and/or (3) the shared assistant device as a whole. Some implementations disclosed herein additionally and/or alternatively relate to: determining, based on utterance(s) of a given user to a shared assistant device, an association of first data and second data; storing the association as accessible to a given user account of the given user; and determining whether to store the association as also accessible by additional user account(s) and/or the shared assistant device.

Abstract: In an implementation, a method for detecting anomalies in textual items is provided. The method includes: receiving a first plurality of textual items by a computing device; training a language model using the received first plurality of textual items by the computing device; after training the language model, receiving a second plurality of textual items by the computing device; calculating a cross-entropy for each textual item in the second plurality of textual items by the computing device using the language model; and detecting an anomaly in at least one of the textual items of the second plurality of textual items by the computing device using the calculated cross-entropies.

Abstract: Systems and processes for operating an intelligent automated assistant are provided. In one example, a plurality of speech inputs is received from a first user. A voice model is obtained based on the plurality of speech inputs. A user input is received from the first user, the user input corresponding to a request to provide access to the voice model. The voice model is provided to a second electronic device.

Abstract: Disclosed is an electronic apparatus which identifies utterer characteristics of an uttered voice input received; identifies one utterer group among a plurality of utterer groups based on the identified utterer characteristics; outputs a recognition result among a plurality of recognition results of the uttered voice input based on a voice recognition model corresponding to the identified utterer group among a plurality of voice recognition models provided corresponding to the plurality of utterer groups, the plurality of recognition results being different in recognition accuracy from one another; identifies recognition success or failure in the uttered voice input with respect to the output recognition result; and changes a recognition accuracy of the output recognition result in the voice recognition model corresponding to the recognition success, based on the identified recognition success in the uttered voice input.

Abstract: An automated speaker verification (ASV) system incorporates a first deep neural network to extract deep acoustic features, such as deep CQCC features, from a received voice sample. The deep acoustic features are processed by a second deep neural network that classifies the deep acoustic features according to a determined likelihood of including a spoofing condition. A binary classifier then classifies the voice sample as being genuine or spoofed.

Abstract: Systems and processes for operating an intelligent automated assistant are provided. In one example, a method includes receiving mixed speech data representing utterances of a target speaker and utterances of one or more interfering audio sources. The method further includes obtaining a target speaker representation, which represents speech characteristics of the target speaker; and determining, using a learning network, probability distributions of phonetic elements directly from the mixed speech data. The inputs of the learning network include the mixed speech data and the target speaker representation. An output of the learning network includes the probability distributions of phonetic elements. The method further includes generating text corresponding to the utterances of the target speaker based on the probability distributions of the phonetic elements; and providing a response to the target speaker based on the text corresponding to the utterances of the target speaker.

Abstract: A processor-implemented method includes: using an encoder, determining, for each of a plurality of tokens included in an input sequence, a self-attention weight based on a token and one or more tokens that precede the token in the input sequence; using the encoder, determining context information corresponding to the input sequence based on the determined self-attention weights; and using a decoder, determining an output sequence corresponding to the input sequence based on the determined context information.

Abstract: Customer support, and other types of activities in which there is a dialogue between two humans can generate large volumes of conversation records. Automated analysis of these records can provide information about high-level features of, for example, the workings of a customer service department. Analysis of these conversations between a customer and a customer-support agent may also allow identification of customer support activities that can be provided by virtual agents instead of actual human agents. The analysis may evaluate conversations in terms of complexity, duration, and sentiment of the participants. Additionally, the conversations may also be analyzed to identify the existence of selected concepts or keywords. Workflow characteristics, the extent to which the conversation represents a multi-step process intended to accomplish a task, may also be determined for the conversations.

Abstract: Computer-implemented systems and methods, trained through machine learning, score a sentiment expressed in a document. Individual sentences are scored and then overall document sentiment score is computed based on scores of individual sentences. Sentence scores can be computed with machine learning models. Digital matrix generator can generate N×M matrix for each sentence, where the matrix comprises vectors of word embeddings for the individual words of the sentence. A classifier computes a sentence sentiment score for each sentence based on the digital matrix for the sentence. Sentence sentiment scores computed by classifier can be adjusted based on a fuzzy matching of a phrase(s) in the sentence to key phrases in a lexicon that are labeled with a sentiment relevant to the context.

Abstract: Disclosed is an electronic device. The electronic device comprises: a microphone comprising circuitry; a speaker comprising circuitry; and a processor electrically connected to the microphone and speaker, wherein the processor, when a first user's voice is input through the microphone, identifies a user who uttered the first user's voice and provides a first response sound, which is obtained by inputting the first user's voice to an artificial intelligence model learned through an artificial intelligence algorithm, through the speaker, and when a second user's voice is input through the microphone, identifies a user who uttered the second user's voice, and if the user who uttered the first user's voice is the same as the user who uttered the second user's voice, provides a second response sound, which is obtained by inputting the second user's voice and utterance history information to the artificial intelligence model, through the speaker.

Abstract: A management of user profiles comprises calculating, for an audio segment, a user confidence measure representing a probability that the audio segment comprises speech of a user and a group confidence measure representing a probability that the audio segment comprises speech of a group of users. A user profile is then managed based on a comparison between the user confidence measure and a user confidence threshold and between the group confidence measure and a group confidence threshold. The embodiments thereby achieve an efficient voice-controlled user profile management by utilizing a layered approach that provides user profiles for group of users as fallback when the identity of the speaking user can not accurately be recognized.

Abstract: Methods and systems for team cooperation with real-time recording of one or more moment-associating elements. For example, a method includes: delivering, in response to an instruction, an invitation to each member of one or more members associated with a workspace; granting, in response to acceptance of the invitation by one or more subscribers of the one or more members, subscription permission to the one or more subscribers; receiving the one or more moment-associating elements; transforming the one or more moment-associating elements into one or more pieces of moment-associating information; and transmitting at least one piece of the one or more pieces of moment-associating information to the one or more subscribers.

Abstract: A device for voice activation includes one or more processors. The one or more processors are configured to receive, via one or more microphones, a keyword and a first command spoken by a first user. The one or more processors are also configured to, subsequent to receiving the first command, receive a second command via the one or more microphones without an intervening receipt of the keyword. The one or more processors are further configured to, based at least in part on determining that the second command is spoken by the same first user, selectively process the second command.

Abstract: Computer-implemented method and system for processing and broadcasting one or more moment-associating elements. For example, the computer-implemented method includes granting subscription permission to one or more subscribers; receiving the one or more moment-associating elements; transforming the one or more moment-associating elements into one or more pieces of moment-associating information; and transmitting at least one piece of the one or more pieces of moment-associating information to the one or more subscribers.