Abstract: A device includes a memory configured to store untransformed ambisonic coefficients at different time segments. The device also includes one or more processors configured to obtain the untransformed ambisonic coefficients at the different time segments, where the untransformed ambisonic coefficients at the different time segments represent a soundfield at the different time segments. The one or more processors are also configured to apply one adaptive network, based on a constraint, to the untransformed ambisonic coefficients at the different time segments to generate transformed ambisonic coefficients at the different time segments, wherein the transformed ambisonic coefficients at the different time segments represent a modified soundfield at the different time segments, that was modified based on the constraint.

Abstract: A method of enhancing an automated speech recognition confidence classifier includes receiving a set of baseline confidence features from one or more decoded words, deriving word embedding confidence features from the baseline confidence features, joining the baseline confidence features with word embedding confidence features to create a feature vector, and executing the confidence classifier to generate a confidence score, wherein the confidence classifier is trained with a set of training examples having labeled features corresponding to the feature vector.

Abstract: A method and apparatus for automatically generating a meeting summary is disclosed herein. Meeting audio is recorded and converted into a text-based transcript. Handwritten meeting notes are converted into notes text. The transcript and notes text are correlated to provide correlated meeting text. Meeting topics are determined from the correlated meeting text. A meeting summary is generated from the meeting topics.

Abstract: The present technique relates to an information processing apparatus and an information processing method that can improve the convenience of a voice AI assistance service used in coordination with content. A first information processing apparatus including an insertion unit that inserts a token, which is related to use of a voice AI assistance service in coordination with content, into an audio stream of the content and a second information processing apparatus including a detection unit that detects the inserted token from the audio stream of the content can be provided to improve the convenience of the voice AI assistance service used in coordination with the content. The present technique can be applied to, for example, a system in coordination with the voice AI assistance service.

Abstract: A method and machine translation system for bi-directional translation of textual sequences between a first language and a second language are described. The machine translation system includes a first autoencoder configured to receive a vector representation of a first textual sequence in the first language and encode the vector representation of the first textual sequence into a first sentence embedding. The machine translation system also includes a sum-product network (SPN) configured to receive the first sentence embedding and generate a second sentence embedding by maximizing a first conditional probability of the second sentence embedding given the first sentence embedding and a second autoencoder receiving the second sentence embedding, the second autoencoder being trained to decode the second sentence embedding into a vector representation of a second textual sequence in the second language.

Abstract: According to one aspect of an embodiment a learning apparatus includes a first acquiring unit that acquires first output information that is output by an output layer when predetermined input information is input to a model that includes an input layer, a plurality of intermediate layers, and the output layer. The learning apparatus includes a second acquiring unit that acquires intermediate output information that is based on pieces of intermediate information that are output by the plurality of intermediate layers when the input information is input to the model. The learning apparatus includes a learning unit that learns the model based on the first output information and the intermediate output information.

Abstract: A method of applying a confidence classifier for intent classification in association with an automated chat bot according to an embodiment includes processing, by a computing system, an utterance with an intent classifier to determine a probability distribution of possible intents associated with the utterance, generating, by the computing system, a plurality of measures of peakedness of the probability distribution, and applying, by the computing system, a trained confidence classifier to determine a single normalized probability of a most likely intent associated with the utterance based on the plurality of measures of peakedness of the probability distribution.

Abstract: An electronic device includes: a camera; a microphone; a display; a memory; and a processor configured to receive an input for activating an intelligent agent service from a user while at least one application is executed, identify context information of the electronic device, control to acquire image information of the user through the camera, based on the identified context information, detect movement of a user's lips included in the acquired image information to recognize a speech of the user, and perform a function corresponding to the recognized speech.

Abstract: Embodiments of the present invention provide a computer system a computer program product, and a method that comprises analyzing a received voice command by identifying a plurality of contextual factors associated with at least one user in a plurality of users using a natural language processing algorithm; dynamically identifying the at least one user in the plurality of users based on an analysis of the identified contextual factors associated with the received voice command; transmitting the received voice command to another computing device within a plurality of computing devices associated with another user in the plurality of users; and generating a line of communication between the plurality of computing devices based on a correlation between a summation of a plurality of security factors and a predetermined threshold of risk associated with authenticating an identity of each user within the plurality of users.

Abstract: Techniques are described for training and/or utilizing an end-to-end speaker diarization model. In various implementations, the model is a recurrent neural network (RNN) model, such as an RNN model that includes at least one memory layer, such as a long short-term memory (LSTM) layer. Audio features of audio data can be applied as input to an end-to-end speaker diarization model trained according to implementations disclosed herein, and the model utilized to process the audio features to generate, as direct output over the model, speaker diarization results. Further, the end-to-end speaker diarization model can be a sequence-to-sequence model, where the sequence can have variable length. Accordingly, the model can be utilized to generate speaker diarization results for any of various length audio segments.

Abstract: In methods and systems for filtering an information input signal, a system may have: a first filter unit filtering an input signal at an initial subinterval in a current update interval according to parameters associated to the preceding update interval, the parameters being scaled by a first scaling factor changing towards 0; and a second filter unit filtering a second filter input signal, based on the output of the first filter unit, at the initial subinterval, according to parameters associated to the current update interval, the parameters being scaled by a second scaling factor changing from 0, or a value close to 0, toward a value more distant from 0.

Abstract: A method of machine learning model development includes building an autoencoder including an encoder trained to map an input into a latent representation, and a decoder trained to map the latent representation to a reconstruction of the input. The method includes building an artificial neural network classifier including the encoder, and a classification layer partially trained to perform a classification in which a class to which the input belongs is predicted based on the latent representation. Neural network inversion is applied to the classification layer to find inverted latent representations within a decision boundary between classes in which a result of the classification is ambiguous, and inverted inputs are obtained from the inverted latent representations. Each inverted input is labeled with a class that is its ground truth, and thereby producing added training data for the classification, and the classification layer is further trained using the added training data.

Abstract: Systems and methods for improving machine learning systems used to model topics on a plurality of calls are described herein. In an embodiment, a server computer receives plurality of digitally stored call transcripts that have been prepared from digitally recorded voice calls. The server computer uses a topic model of an artificial intelligence machine learning system, the topic model modeling words of a call as a function of one or more word distributions for each topic of a plurality of topics, to generate an output of the topic model which identifies the plurality of topics represented in the plurality of call transcripts. The server computer computes, for a particular topic of the plurality of topics a first value representing a vocabulary of the particular topic and a second value representing a consistency of the particular topic in two more call transcripts of the plurality of call transcripts which include the particular topic.

Abstract: Techniques are disclosed for automatically adjusting machine learning parameters in an e-commerce system. Hyperparameters of a machine learning component are tuned using a gradient estimator and a first training set representative of an e-commerce context. The machine learning component is trained using the tuned hyperparameters and the first training set. The hyperparameters are automatically re-tuned using the gradient estimator and a second training set representative of a changed e-commerce context. The machine learning component is re-trained using the re-tuned hyperparameters and the second training set.

Abstract: A device implementing a system for processing speech in an audio signal includes at least one processor configured to receive an audio signal corresponding to at least one microphone of a device, and to determine, using a first model, a first probability that a speech source is present in the audio signal. The at least one processor is further configured to determine, using a second model, a second probability that an estimated location of a source of the audio signal corresponds to an expected position of a user of the device, and to determine a likelihood that the audio signal corresponds to the user of the device based on the first and second probabilities.

Abstract: The present teaching relates to method, system, medium, and implementation for activating an animatronic device. Information about a user is obtained for whom an animatronic device is to be configured to carry out a dialogue with the user. The animatronic device includes a head portion and a body portion and the head portion is configured based on one of a plurality of selectable head portions. One or more preferences of the user are identified from the obtained information and used to select, from the plurality of selectable head portions, a first selected head portion. A configuration of the head portion of the animatronic device is then configured based on the first selected head portion for carrying out the dialogue.

Abstract: Embodiments relate to a system, program product, and method for leveraging cognitive systems to facilitate the management of word and image relationships in a combined vector space. More specifically, the system, computer program product, and method disclosed herein facilitate establishing one or more continuous semantic relationships with word vectors, sentence vectors, and image vectors within an image-word combined vector space with N-dimensional coordinates. In general, the word vectors, sentence vectors, and image vectors are generated within their respective domains that are resident within the image-word combined vector space. Such domains of word vectors, sentence vectors, and image vectors include inherent semantic relationships between respective members of each domain that facilitate establishing continuous, formal, virtual relationships between respective word vectors, image vectors, and sentence vectors within the image-word combined vector space.

Abstract: An encoding method includes determining a target adaptive broadening factor based on a quantized line spectral frequency (LSF) parameter of a primary channel signal in a current frame and an LSF parameter of a secondary channel signal in the current frame, and writing the quantized LSF parameter of the primary channel signal in the current frame and the target adaptive broadening factor into a bitstream.

Abstract: Systems and methods for suppressing noise and detecting voice input in a multi-channel audio signal captured by two or more network microphone devices include receiving an instruction to process one or more audio signals captured by a first network microphone device and after receiving the instruction (i) disabling at least a first microphone of a plurality of microphones of a second network microphone device, (ii) capturing a first audio signal via a second microphone of the plurality of microphones, (iii) receiving over a network interface of the second network microphone device a second audio signal captured via at least a third microphone of the first network microphone device, (iv) using estimated noise content to suppress first and second noise content in the first and second audio signals, (v) combining the suppressed first and second audio signals into a third audio signal, and (vi) determining that the third audio signal includes a voice input comprising a wake word.

Abstract: Technologies are disclosed for interacting with a virtual assistant to request updates associated with one or more events and/or perform actions. According to some examples, a user may use their voice to interact with a virtual assistant to receive updates relating to events occurring during a certain period of time. For example, a user may request an update associated with one or more events occurring that day. The system may access data sources (e.g., calendar services, email services, etc.) to obtain data associated with the events, tag the events according to one or more conditions indicated by the data, and/or rank the events according to the tags. In addition, to resolve conditions associated with the events, the virtual assistant may also include options in the update to perform certain actions and/or to provide response data. The virtual assistant may generate the update and audibly provide the update to the user.