Abstract: Agents engage and disengage with users intelligently. Users can tell agents to remain engaged without requiring a wakeword. Engaged states can support modal dialogs and barge-in. Users can cause disengagement explicitly. Disengagement can be conditional based on timeout, change of user, or environmental conditions. Engagement can be one-time or recurrent. Recurrent states can be attentive or locked. Locked states can be unconditional or conditional, including being reserved to support user continuity. User continuity can be tested by matching parameters or tracking user by many modalities including microphone arrays, cameras, and other sensors.

Abstract: An artificial intelligence voice interaction method and a near-end electronic device thereof are disclosed. The method includes the following steps: receiving a voice input by a user; transmitting the voice to a remote artificial intelligence server; determining whether the voice has ended; when determining that the voice has ended and has not received a stop recording signal transmitted by the remote artificial intelligence server, it stops transmitting the voice to the remote artificial intelligence server; before determining that the voice has ended, and has received the stop recording signal from the remote artificial intelligence server, it stops transmitting the voice to the remote artificial intelligence server; and receiving a response signal send back from the remote artificial intelligence server.

Abstract: A method, performed by a device, of providing a response to a user's voice input, includes capturing, via a camera of the device, an image including at least one object; activating a microphone of the device as the image is captured; receiving, via the microphone, the user's voice input for the object; determining the intention of the user with respect to the object by analyzing the received voice input; and providing a response regarding the at least one object based on the determined intention of the user.

Abstract: A server may provide a voice recognition service. The server may include a memory configured for storing a plurality of voice recognition models, a communication device configured for communicating a plurality of voice recognition devices, and an artificial intelligence device configured for providing a voice recognition service to the plurality of voice recognition devices, acquiring use-related information regarding a first voice recognition device (from among the plurality of voice recognition devices), and changing a voice recognition model corresponding to the first voice recognition device from a first voice recognition model to a second voice recognition model based on the use-related information.

Abstract: There is provided an information processing device to achieve more flexible correction of a recognized sentence, the information processing device including: a comparison unit configured to compare first sound-related information obtained from collected first utterance information with second sound-related information obtained from collected second utterance information; and a setting unit configured to set a new delimiter position different from a result of speech-to-text conversion associated with the first utterance information on a basis of a comparison result obtained by the comparison unit.

Abstract: A hearing aid and a hearing aid output voice adjustment method thereof are disclosed. The hearing aid includes a microphone, a voice output adjustment module, a voice player, and a voice processing module. The microphone is used for receiving an input voice. The voice player is used for outputting an output voice. The voice processing module is used for receiving the input voice and modifying to the output voice, wherein the voice processing module is used for adjusting sound output energies of at least N steps, wherein the sound output energy of the nth step is greater than that of the n?1th step, 2?n?N, N?3. When the voice processing module adjusts the sound output energy of the Nth step, the voice output adjustment module also simultaneously reduces at least a portion of frequencies of the input voice.

Abstract: One or more implementations of the present specification provide a social content risk identification method. Social content data to be identified is obtained. Features of the social content data are extracted, including a plurality of features of at least one of social behavior records or social message records in the social content data. The features are expanded by generating dimension-extended features using a tree structured machine learning model. The social content data is classified as risky social content data by processing the dimension-extended features using a deep machine learning model.

Abstract: Described systems and methods allow an automatic translation from a natural language (e.g., English) into an artificial language such as a structured query language (SQL). In some embodiments, a translator module includes an encoder component and a decoder component, both components comprising recurrent neural networks. Training the translator module comprises two stages. A first stage trains the translator module to produce artificial language (AL) output when presented with an AL input. For instance, the translator is first trained to reproduce an AL input. A second stage of training comprises training the translator to produce AL output when presented with a natural language (NL) input.

Abstract: Disclosed is an artificial intelligence voice recognition apparatus including: a microphone configured to receive a voice command; a memory configured to store a first voice recognition algorithm; a communication module configured to transmit the voice command to a server system and receive first voice recognition algorithm-related update data from the server system; and a controller configured to perform control to update the first voice recognition algorithm, which is stored in the memory, based on the first voice recognition algorithm-related update data. Accordingly, the voice recognition apparatus is able to provide a voice recognition algorithm fitting to a user's characteristics.

Abstract: In at least one broad aspect, described herein are systems and methods in which a latent representation shared between two languages is built and/or accessed, and then leveraged for the purpose of text generation in both languages. Neural text generation techniques are applied to facilitate text generation, and in particular the generation of sentences (i.e., sequences of words or subwords) in both languages, in at least some embodiments.

Abstract: Examples are disclosed that relate to voice identification enrollment. One example provides a method of voice identification enrollment comprising, during a meeting in which two or more human speakers speak at different times, determining whether one or more conditions of a protocol for sampling meeting audio used to establish human speaker voiceprints are satisfied, and in response to determining that the one or more conditions are satisfied, selecting a sample of meeting audio according to the protocol, the sample representing an utterance made by one of the human speakers. The method further comprises establishing, based at least on the sample, a voiceprint of the human speaker.

Abstract: An image is received from a light capture device associated with the smart display device. A determination is made as to whether to activate voice recognition of a recording device associated with the smart display device based on a face being in the image. In response to determining to activate the voice recognition of the recording device associated with the smart display device based on the face being in the image, the voice recognition of the recording device associated with the smart display device is activated.

Abstract: Disclosed are a response device determination method and a response device determination apparatus. The method includes receiving audio signals from a plurality of devices respectively; extracting a plurality of distance information indicative of distances between the user and the plurality of devices from the audio signals respectively; and determining a response device to respond to the wake-up voice using the extracted plurality of distance information, wherein the response device is determined based on at least one of first and second steps according to a predetermined condition, wherein the first step includes comparing the extracted plurality of distance information with each other and determining the response device based on the comparison result, wherein the second step includes applying the extracted plurality of distance information to a deep neural network (DNN) model to obtain an application result and determining the response device based on the application result.

Abstract: Detecting and monitoring legacy devices (such as appliances in a home) using audio sensing is disclosed. Methods and systems are provided for transforming audio data captured by the sensor to afford privacy when speech is overheard by the sensor. Because these transformations may negatively impact the ability to detect/monitor devices, an effective transformation is determined based on both privacy and detectability concerns.

Abstract: A speaker estimation method that estimate the speaker from audio and image includes: inputting audio; extracting a feature quantity representing a voice characteristic from the input audio; inputting an image; detecting person regions of respective persons from the input image; estimating feature quantities representing voice characteristics from the respective detected person regions; Performing a change such that an image taken from another position and with another angle is input when any person is not detected; calculating a similarity between the feature quantity representing the voice characteristic extracted from the audio and the feature quantity representing the voice characteristic estimated from the person region in the image; and estimating a speaker from the calculated similarity.

Abstract: One embodiment provides a method, comprising: capturing, at an information handling device, audible input from at least one user; providing, on a display device operatively coupled to the information handling device, at least one transcription suggestion, wherein the at least one transcription suggestion is associated with a portion of the audible input; and inputting, responsive to receiving a selection input on the at least one transcription suggestion, the at least one transcription suggestion into an underlying application. Other aspects are described and claimed.

Abstract: A system and method of establishing a communication session is disclosed herein. A computing system receives, from a client device, a content item comprising text-based content. The computing system generates a mark-up version of the content item by identifying one or more characters in the text-based content and a relative location of the one or more characters in the content item. The computing system receives, from the client device, an interrogatory related to the content item. The computing system analyzes the mark-up version of the content item to identify an answer to the interrogatory. The computing system generates a response message comprising the identified answer to the interrogatory. The computing system transmits the response message to the client device.

Abstract: According to one embodiment, a dialog system includes an acquirer and a processor. The acquirer acquires a first sentence. The processor refers to a first weighting amount based on a first feature and a first weight. The first feature is of a first dialog state and relates to a first slot. The first weight pertains to the first sentence. The processor refers to a second weighting amount based on a first intention amount and a second weight. The first intention amount is of a first intention and relates to the first slot. The first intention is included in the first sentence. The second weight pertains to the first sentence and is different from the first weight. The processor outputs a second feature using the first and second weighting amounts. The second feature is of a second dialog state after the first dialog state and relates to the first slot.

Abstract: Electronic records are accessed from computer storage for a given subject, wherein the electronic records include natural language notes about the subject. Tokens are identified in the natural language notes. For each token, a corresponding intensity score is generated representing an intensity of match between the token and a particular dimension, wherein the intensity scores are each values on a first scale, wherein each dimension is one of a plurality of dimensions of a category out of a plurality of categories; generating rescaled-intensity scores from the intensity scores by rescaling the intensity scores from the first scale to a second scale different from the first scale. For each dimension of each category, a dimension-score is compiled based on the intensity scores; and categorizing the subject into at least one category based on the dimension scores. The subject is categorized into at least one category based on the dimension scores.

Abstract: Methods, systems, and related products that provide emotion-sensitive responses to user's commands and other utterances received at an utterance-based user interface. Acknowledgements of user's utterances are adapted to the user and/or the user device, and emotions detected in the user's utterance that have been mapped from one or more emotion features extracted from the utterance. In some examples, extraction of a user's changing emotion during a sequence of interactions is used to generate a response to a user's uttered command. In some examples, emotion processing and command processing of natural utterances are performed asynchronously.