Abstract: Embodiments described herein provide for systems and methods for implementing a neural network architecture for spoof detection in audio signals. The neural network architecture contains a layers defining embedding extractors that extract embeddings from input audio signals. Spoofprint embeddings are generated for particular system enrollees to detect attempts to spoof the enrollee's voice. Optionally, voiceprint embeddings are generated for the system enrollees to recognize the enrollee's voice. The voiceprints are extracted using features related to the enrollee's voice. The spoofprints are extracted using features related to features of how the enrollee speaks and other artifacts. The spoofprints facilitate detection of efforts to fool voice biometrics using synthesized speech (e.g., deepfakes) that spoof and emulate the enrollee's voice.

Abstract: One or more computing devices, systems, and/or methods for user identification and authorization are provided. In an example, a voice command to perform an action is detected. A voice profile associated with a user is identified based upon voice characteristics of the voice command. In response to determining that the voice profile is not linked to an account associated with the action, the user is prompted for an identifier associated with a device for creating the account through the device. In response to receiving the identifier from the user, the identifier is utilized to facilitate creation of the account through the device.

Abstract: Techniques for using validated communications identifiers of a user's communications profile to resolve entries in another user's contact list are described. When a user imports a contact list, the contact list may include multiple entities related to the same person. The system may identify one of the entries in the contact list that corresponds to a validated communications identifier stored in another user's communications profile. The system may identify other validated communications identifiers in the other user's communications profile and cross-reference them against the entries of the contact list. If the system determines the contact list includes entries for the different validated communications identifiers of the other user, the system may consolidate the entries into a single entry associated with the other user.

Abstract: A feature vector having high class identification capability is generated. A signal processing system provided with: a first generation unit for generating a first feature vector on the basis of one of time-series voice data, meteorological data, sensor data, and text data, or on the basis of a feature quantity of one of these; a weight calculation unit for calculating a weight for the first feature vector; a statistical amount calculation unit for calculating a weighted average vector and a weighted high-order statistical vector of second or higher order using the first feature vector and the weight; and a second generation unit for generating a second feature vector using the weighted high-order statistical vector.

Abstract: Implementations are directed to biasing speaker authentication on a per-user basis and on a device-by-device basis and/or contextual feature(s) basis. In some of those implementations, in performing speaker authentication based on a spoken utterance, different biasing parameters are determined for each of multiple different registered users of an assistant device at which the spoken utterance was detected. In those implementations, each of the biasing parameters can be used to make it more likely or less likely (in dependence of the biasing parameter) that a corresponding registered user will be verified using the speaker authentication. Through utilization of biasing parameter(s) in performing speaker authentication, accuracy and/or robustness of speaker authentication can be increased.

Abstract: Methods and apparatuses are described for integrating common and context-specific natural language understanding (NLU) processing in a virtual assistant application. A server identifies a common NLU processing unit comprising (i) a common NLU classification model and (ii) a common response library comprising common responses. The server builds a combined NLU processing unit based upon the common NLU classification model and a context-specific NLU classification model, the combined NLU processing unit comprising (i) a combined NLU classification model trained using the common NLU training data and context-specific NLU training data, (ii) the common response library and (iii) a context-specific response library comprising context-specific responses. The server generates a response to an input utterance by determining a user intent associated with the input utterance using the combined NLU classification model, locating a response associated with the user intent in one or more of the libraries.

Abstract: A method includes receiving acoustic features of a first utterance spoken by a first user that speaks with typical speech and processing the acoustic features of the first utterance using a general speech recognizer to generate a first transcription of the first utterance. The operations also include analyzing the first transcription of the first utterance to identify one or more bias terms in the first transcription and biasing the alternative speech recognizer on the one or more bias terms identified in the first transcription. The operations also include receiving acoustic features of a second utterance spoken by a second user that speaks with atypical speech and processing, using the alternative speech recognizer biased on the one or more terms identified in the first transcription, the acoustic features of the second utterance to generate a second transcription of the second utterance.

Abstract: Aspects of the disclosure provide for mechanisms for identification of fields in documents using neural networks. A method of the disclosure includes obtaining a layout of a document, the document having a plurality of fields, identifying the document, based on the layout, as belonging to a first type of documents of a plurality of identified types of documents, identifying a plurality of symbol sequences of the document, and processing, by a processing device, the plurality of symbol sequences of the document using a first neural network associated with the first type of documents to determine an association of a first field of the plurality of fields with a first symbol sequence of the plurality of symbol sequences of the document.

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: The present disclosure relates to a speaker model adaptation method and device for enhancing text-independent speaker recognition performance. Specifically, the disclosure relates to a method and a device whereby, for the adaption of a speaker model pre-stored in an electronic device, text-independent speaker recognition performance is improved by considering variations in the amount of speaker characteristics information per phoneme unit.

Abstract: Systems, methods, and non-transitory computer-readable media can obtain a stream of audio waveform data that represents speech involving a plurality of speakers. As the stream of audio waveform data is obtained, a plurality of audio chunks can be determined. An audio chunk can be associated with one or more identity embeddings. The stream of audio waveform data can be segmented into a plurality of segments based on the plurality of audio chunks and respective identity embeddings associated with the plurality of audio chunks. A segment can be associated with a speaker included in the plurality of speakers. Information describing the plurality of segments associated with the stream of audio waveform data can be provided.

Abstract: A computer-implemented system, method, and computer program product are described that predicts a time in which a participant of a group of participants may speak in a group meeting. The prediction is based in part on latency between each participant in the group of participants and learned voice patterns of each respective participant. Once a prediction is determined, a cue to speak or remain silent is provided as one of a visual or aural cue to the participant to speak during the group meeting. Predicting a time in which the participant of a group of participants may speak includes training a lingual processing model iteratively for each respective participant to learn voice patterns of participants.

Abstract: A device includes one or more processors configured to receive an audio data sample and to provide the audio data sample to a dynamic classifier. The dynamic classifier is configured to generate a classification output corresponding to the audio data sample. The one or more processors are further configured to selectively access a particular device based on the classification output.

Abstract: Interactive natural language processing (NLP) results may be generated that allow a user to interact with the NLP results but do so in an offline manner so that the documents being processed need not be stored online. To provide interactive NLP results, event handlers may be attached to elements of the NLP results. A user may then select a word or phrase of the NLP results to cause computer software provided with the NLP to present the interactive features. For example, a user may click on a definite noun phrase to view information for diagnosing antecedent basis errors. For another example, a user may click on a word to view information about how that word is used in a document, such as viewing portions of the document that include the word or variants of the word.

Abstract: Provided are computer implemented systems and methods for providing improved language translation, including translation of input text to an output translated text element. This may include providing for rule-based language translation. This may further include a plurality of machine translation models adapted to different translation domains. A machine translation selector is described that selects an in-domain machine translator from the plurality of machine translation models based on an input text element. The selected in-domain machine translator may generate a first translated text element from the input text element. A post-editor module may generate a second translated text element by predicting a post-edit to the first translated text element. The first translated text element and the second translated text element are evaluated by a quality evaluation model that determines a first metric associated with the first translated text element and a second metric associated with the second text element.

Abstract: Aspects relate to apparatuses and methods for integrated application tracking. An exemplary apparatus includes a cloud platform, a processor, and a memory communicatively connected to the processor, the memory containing instructions configuring the processor to receive a plurality of user data related to a user, parse the plurality of user data into a key work record, generate a user metric, as a function of the key work record, based on a plurality of weighted values reflecting desirability, wherein generating the user metric includes identifying at least a negative factor in the key work record, and adjusting the plurality of weighted values as a function of the at least a negative factor and a temporal displacement of the at least a negative factor.

Abstract: A computing device may receive an indication of an audio signal captured by a microphone, wherein the audio signal includes voice input. The computing device may determine that the voice input in the audio signal is from an authorized user of the computing device and includes a trigger phrase associated with a request to trigger device finder functionality based at least in part on comparing the voice input with data provided by the authorized user of the computing device. The computing device may, in response to determining that the voice input in the audio signal is from the authorized user of the computing device and includes the trigger phrase associated with the request to trigger device finder functionality, cause a speaker of the computing device to audibly output the alert sound to assist the authorized user to locate the computing device.

Abstract: The invention provides a computer implemented method of drafting of abbreviations for the statistically most frequent word forms and phrases for the purposes of computer typing and compression of texts written in languages using alphabetic scripts with full vowel representation. Therein, drafted abbreviations do not constitute meaningful word forms of a given language, for which they are drafted. Every abbreviated word form or phrase is attributed only one unique and exclusive abbreviation, which is based on the letters contained in this abbreviated word form or phrase and in accordance with the order, in which these letters appear in the abbreviated word form or phrase. For a given word form, one-letter, two-letter, three-letter and four-letter abbreviations of the word forms are chosen according to the statistical frequency of the word forms in a way that allows the mathematically most efficient process of abbreviation of the text.

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 system for capturing and processing portions of a spoken utterance command that may occur before a wakeword. The system buffers incoming audio and indicates locations in the audio where the utterance changes, for example when a long pause is detected. When the system detects a wakeword within a particular utterance, the system determines the most recent utterance change location prior to the wakeword and sends the audio from that location to the end of the command utterance to a server for further speech processing.