Abstract: A system is configured to: (a) ingest a set of codes from a code reference, the code reference including words and non-words; (b) associate the set of codes to a set of representations, forming descriptive links between the set of codes and the set of representations; (c) ingest an unstructured document having a general outline, the unstructured document including (i) at least one of the words in the code reference, (ii) at least one code in the set of codes, or (iii) both (i) and (ii); (d) determine an ontology from content of the unstructured document with an unsupervised learning algorithm; and (e) store the determined ontology in a database, the ontology having an acyclic directed graph structure.

Abstract: A speech recognition method includes receiving a first multi-channel audio signal; obtaining at least one of a speech signal characteristic or a noise signal characteristic for at least one frequency band of frequency bands corresponding to channel audio signals included in the first multi-channel audio signal; generating a signal with an enhanced speech component by performing beamforming on the first multi-channel audio signal based on the speech signal characteristic, a speech signal characteristic obtained for a previous frame that was obtained before a certain time that the first multi-channel audio signal was obtained, and the noise signal characteristic; determining whether the enhanced speech component includes a wake word; and based on determining that the enhanced speech component includes the wake word: activating a speech recognition operation based on the signal with the enhanced speech component.

Abstract: The present document relates to a method of layered encoding of a frame of a compressed higher-order Ambisonics, HOA, representation of a sound or sound field. The compressed HOA representation comprises a plurality of transport signals. The method comprises assigning the plurality of transport signals to a plurality of hierarchical layers, the plurality of layers including a base layer and one or more hierarchical enhancement layers, generating, for each layer, a respective HOA extension payload including side information for parametrically enhancing a reconstructed HOA representation obtainable from the transport signals assigned to the respective layer and any layers lower than the respective layer, assigning the generated HOA extension payloads to their respective layers, and signaling the generated HOA extension payloads in an output bitstream.

Abstract: The present document relates to a method of layered encoding of a compressed sound representation of a sound or sound field. The compressed sound representation comprises a basic compressed sound representation comprising a plurality of components, basic side information for decoding the basic compressed sound representation to a basic reconstructed sound representation of the sound or sound field, and enhancement side information including parameters for improving the basic reconstructed sound representation.

Abstract: A remote personal assistant server/service is configured with a plurality of different state machines for supporting asynchronous communications with the front-end interfaces of personal assistants and for supporting the interfacing with a plurality of proprietary back-end intelligence bots, which are independently controlled and separate from the remote personal assistant service/server.

Abstract: A system, method, and computer program product for determining statement attributions. The system includes at least one processing component, at least one memory component, a feature extractor, a model generator, a model database, and an attribution selector. The method includes receiving a statement, generating at least one pattern that defines a grammatical feature of the statement, and generating a statement model from the at least one pattern. The method also includes determining a similarity value for the statement model and at least one reference model.

Abstract: The present disclosure discloses a data processing method and apparatus for an intelligent device, and a storage medium, which relates to a field of artificial intelligence technologies. The method includes: extracting key voice information from collected user voice information; in a non-wireless fidelity (WiFi) network environment, transmitting the key voice information to a mobile terminal, so that the mobile terminal transmits the key voice information to a server, and receives a processing result fed back by the server after the server processes the key voice information; and obtaining the processing result from the mobile terminal to display the processing result.

Abstract: A method includes obtaining audio data for a long-form utterance and segmenting the audio data for the long-form utterance into a plurality of overlapping segments. The method also includes, for each overlapping segment of the plurality of overlapping segments: providing features indicative of acoustic characteristics of the long-form utterance represented by the corresponding overlapping segment as input to an encoder neural network; processing an output of the encoder neural network using an attender neural network to generate a context vector; and generating word elements using the context vector and a decoder neural network. The method also includes generating a transcription for the long-form utterance by merging the word elements from the plurality of overlapping segments and providing the transcription as an output of the automated speech recognition system.

Abstract: Methods and apparatuses are described for building a knowledge base taxonomy from structured or unstructured computer text for use in automated user interactions. A server computing device receives one or more of structured text or unstructured text corresponding to historical user interaction data from a database. The server computing device extracts one or more terms from the received text that are most relevant to a subject matter domain. The server computing device organizes the extracted one or more terms into a taxonomy data structure.

Abstract: This system includes: an apparatus; a first server; and a second server connectable to the apparatus and the first server. The first server includes a voice receiver that receives a voice from a user, a command transmitter that transmits a command based on the voice to the second server, and an output part that, responsive to a response signal received from the second server, provides an output based on the response signal. The second server includes a command processor that transmits a response signal to the first server responsive to a first command received from the first server, and that acquires a status of the apparatus, and a status transmitter that transmits a response signal based on the acquired status of the apparatus to the first server responsive to a second command received from the first server.

Abstract: When the noise in an audio signal made up of both speech and noise is suppressed, the quality of the speech in the audio signal is usually degraded. The speech obtained from a noise-suppressed audio signal is improved by determining linear predictive coding (LPC) characteristics of the audio signal without or prior to noise suppression and by determining the LPC characteristics of the noise-suppressed audio. The convolution of those differing characteristics provides an improved-quality speech signal, with the original noise level reduced or suppressed.

Abstract: This disclosure proposes systems and methods enabling on-device/hybrid processing of speech requests using a hub device. The hub device is capable of receiving audio data from surrounding devices and performing speech processing on the audio data to improve latency and/or provide functionality to other devices within a private network. The hub device may receive multiple requests corresponding to different utterances. If the hub device receives a second utterance while processing a first utterance, the hub device may send an error notification, process the first utterance and the second utterance sequentially, suspend processing of the first utterance to process the second utterance first, send the second utterance to another hub device or remote system, or suspend processing of the first utterance and send the first utterance to the remote system in order to process the second utterance.

Abstract: Systems and methods for processing speech are described. In certain examples, image data is used to generate visual feature tensors and audio data is used to generate audio feature tensors. The visual feature tensors and the audio feature tensors are used by a linguistic model to determine linguistic features that are usable to parse an utterance of a user. The generation of the feature tensors may be jointly configured with the linguistic model. Systems may be provided in a client-server architecture.

Abstract: In an acoustic signal separation device (1), a determination unit (6) determines whether or not components from a plurality of sound sources are mixed in each of acoustic signals of respective components regenerated by a signal regeneration unit (5), and when it is determined that a plurality of components is mixed, a series of processes by a feature value extraction unit (2), a data estimation unit (3), a data classification unit (4), and a signal regeneration unit (5) is repeatedly executed until acoustic signals of the components of the respective sound sources are regenerated.

Abstract: The technology described in this document can be embodied in a computer-implemented method that includes presenting a first user-interface that includes a user-selectable menu of multiple virtual personal assistant (VPA) service providers, and receiving a user-selection identifying a particular VPA service provider. The method also includes receiving a first signal representing input speech data, and processing the first signal to generate a first electronic file that includes at least a portion of the input speech data, the first electronic file being generated in accordance with a specification of the particular VPA service provider. The method further includes transmitting the first electronic file to one or more remote computing devices associated with the particular VPA service provider, receiving at least a second electronic file including a response to the input speech data, and causing an acoustic transducer to generate an acoustic output based on the second electronic file.

Abstract: Some implementations are directed to translating chatspeak to a normalized form, where the chatspeak is included in natural language input formulated by a user via a user interface input device of a computing device—such as input provided by the user to an automated assistant. The normalized form of the chatspeak may be utilized by the automated assistant in determining reply content that is responsive to the natural language input, and that reply content may be presented to the user via one or more user interface output devices of the computing device of the user. Some implementations are additionally and/or alternatively directed to providing, for presentation to a user, natural language output that includes chatspeak in lieu of a normalized form of the chatspeak, based at least in part on a “chatspeak measure” that is determined based on past usage of chatspeak by the user and/or by additional users.

Abstract: Techniques for speech recognition are described. For example, a speech recognition service is to receive a request to perform speech recognition on speech data from a chatbot using a particular speech recognition service; determine a group of hosts to route the speech data to, the group of hosts to host a plurality of speech recognition services including the particular speech recognition service; determine a path to the determined group of hosts using a set of one or more rules; determine a particular host of the group of hosts to perform speech recognition on the speech data, the particular host having the speech recognition service in memory to process the request and being preferred for performing the speech recognition on the speech data; route the speech data to the particular host; perform speech recognition on the speech data using the particular host; and provide a text result of the speech recognition.

Abstract: Systems and processes for operating an intelligent automated assistant to perform intelligent list reading are provided. In one example process, a spoken user request associated with a plurality of data items is received. The process determines whether a degree of specificity of the spoken user request is less than a threshold level. In response to determining that a degree of specificity of the spoken user request is less than a threshold level, one or more attributes related to the spoken user request are determined. The one or more attributes are not defined in the spoken user request. Additionally, a list of data items based on the spoken user request and the one or more attributes is obtained. A spoken response comprising a subset of the list of data items is generated and the spoken response is provided.

Abstract: A laundry scheduling device according to an embodiment of the present invention includes an input interface that receives speech including a plurality of words respectively representing a plurality of laundry items from a use, and a processor that acquires one or more features represented by each of the plurality of words by inputting speech data corresponding to the received speech to a learning model and performing word embedding, and generates a laundry schedule of the plurality of laundry items based on the one or more features represented by each of the plurality of words.

Abstract: This disclosure relates generally to speech signal processing, and more particularly to a method and system for processing speech signal for emotion identification. The system processes a speech signal collected as input, during which a plurality of differential features corresponding to a plurality of frames of the speech signal are extracted. Further, the differential features are compared with an emotion recognition model to identify at least one emotion matching the speech signal, and then the at least one emotion is associated with the speech signal.