Abstract: Example aspects include techniques for implementing real-time and low-latency synthesis of audio. These techniques may include generating a frame by sampling audio input in increments equal to a buffer size of until a threshold corresponding to a frame size used to train a machine learning (ML) model is reached, detecting feature information within the frame, determining, by the ML model, control information for audio reproduction based on the feature information. In addition, the techniques may include generating filtered noise information by inverting the noise magnitude control information using an overlap and add technique, generating, based on the control information, additive harmonic information by combining a plurality of scaled wavetables, and rendering audio output based on the filtered noise information and the additive harmonic information.

Abstract: One or more user interactions directed to a set of one or more voice-controlled devices in an environment are received by a first connected device. A first input to a first voice-controlled device of the set of voice-controlled devices is detected based on the user interactions. A potential second input to the set of voice-controlled devices is determined in response to the first input and based on an activity model. A deviation from the potential second input is monitored for, in response to the first input and from the user interactions. An activity anomaly in the environment is identified based on the monitoring. A correction action is performed in response to the activity anomaly.

Abstract: One or more systems, devices, computer program products and/or computer-implemented methods provided herein relate to determining veracity of answers generated by machine comprehension question and answer models. According to an embodiment, a machine comprehension component can generate a first answer to a query by extracting the first answer from a passage of text corpus. The text corpus alteration component can alter the text corpus one or more times to produce one or more altered text corpora. The machine comprehension component can further extract one or more additional answers to the query from the altered text corpora. A comparison component can determine a veracity score for the first answer based on one or more comparisons of the first answer with the one or more additional answers.

Abstract: Techniques are disclosed herein relating to using reinforcement learning to generate a dialogue policy. A computer system may perform an iterative training operation to train a deep Q-learning network (DQN) based on conversation logs from prior conversations. In various embodiments, the DQN may include an input layer to receive an input value indicative of a current state of a given conversation, one or more hidden layers, and an output layer that includes a set of nodes corresponding to available responses. During the iterative training operation, the disclosed techniques may analyze utterances from a conversation log and, based on the utterances, use the DQN to determine appropriate responses. Reward values may be determined based on the selected responses and, based on the reward values, the DQN may be updated. Once generated, the dialogue policy may be used by a chatbot system to guide conversations with users.

Abstract: Systems and methods receive, in real-time from a user via a user device, input audio data comprising communication element(s) and trained model(s) are applied thereto to categorize the communication element(s), the categorizing comprising assigning a contextual category to a communication element. Text is generated that includes a response to the communication element(s), the response including individualized and contextualized qualities predicted to provide an optimal outcome based on (i) the assigned contextual category and (ii) the user. Text-to-speech processing of the text is implemented to produce an audio output comprising (a) the response and (b) a speech pattern predicted to facilitate the optimal outcome. The audio output is provided to the user via the user device, and based thereon a user's reaction is measured according to a quantifiable quality score that is used to modify future iterations of text-to-speech processing to provide future audio output(s) including a revised speech pattern.

Abstract: Systems, methods, and devices for human-machine interfaces for utterance-based playlist selection are disclosed. In one method, a list of playlists is traversed and a portion of each is audibly output until a playlist command is received. Based on the playlist command, the traversing is stopped and a playlist is selected for playback. In examples, the list of playlists is modified based on a modification input.

Abstract: This disclosure relates generally to extraction of cause-effect relation from domain specific text. Cause-effect relation highlights causal relationship among various entities, concepts and processes in a domain specific text. Conventional state-of-the-art methods use named entity recognition for extraction of cause-effect (CE) relation which does not give precise results. Embodiments of the present disclosure provide a knowledge-based approach for automatic extraction of CE relations from domain specific text. The present disclosure method is a combination of an unsupervised machine learning technique to discover causal triggers and a set of high-precision linguistic rules to identify cause/effect arguments of these causal triggers. The method extracts the CE relation in the form of a triplet comprising a causal trigger, a cause phrase and an effect phrase identified from the domain specific text. The disclosed method is used for extracting CE relations in biomedical text.

Abstract: A system and a method are disclosed for identifying a subjectively interesting moment in a transcript. In an embodiment, a device receives a transcription of a conversation, and identifies a participant of the conversation. The device accesses a machine learning model corresponding to the participant, and applies, as input to the machine learning model, the transcription. The device receives as output from the machine learning model a portion of the transcription having relevance to the participant, and generates for display, to the participant, information pertaining to the portion.

Abstract: Eyewear having a speech to moving lips algorithm that receives and translates speech and utterances of a person viewed through the eyewear, and then displays an overlay of moving lips corresponding to the speech and utterances on a mask of the viewed person. A database having text to moving lips information is utilized to translate the speech and generate the moving lips in near-real time with little latency. This translation provides the deaf/hearing impaired users the ability to understand and communicate with the person viewed through the eyewear when they are wearing a mask. The translation may include automatic speech recognition (ASR) and natural language understanding (NLU) as a sound recognition engine.

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.

Abstract: An apparatus for generating a lip sync image according to disclosed embodiment has one or more processors and a memory which stores one or more programs executed by the one or more processors. The apparatus includes a first artificial neural network model configured to generate an utterance match synthesis image by using a person background image and an utterance match audio signal corresponding to the person background image as an input, and generate an utterance mismatch synthesis image by using the person background image and an utterance mismatch audio signal not corresponding to the person background image as an input, and a second artificial neural network model configured to output classification values for an input pair in which an image and a voice match and an input pair in which an image and a voice do not match by using the input pairs as an input.

Abstract: The present disclosure discloses a question analysis method, a device, a knowledge base question answering system and an electronic equipment. The method includes: analyzing a question to obtain N linearized sequences, N being an integer greater than 1; converting the N linearized sequences into N network topology maps; separately calculating a semantic matching degree of each of the N network topology maps to the question; and selecting a network topology map having a highest semantic matching degree to the question as a query graph of the question from the N network topology maps. According to the technology of the present disclosure, the query graph of the question can be obtained more accurately, and the accuracy of the question to the query graph is improved, thereby improving the accuracy of question analysis.

Abstract: Methods, systems, and computer program products for detecting contextual bias in text are provided herein. A computer-implemented method includes identifying, by a machine learning network, a protected attribute in one or more data samples; processing the identified data samples using a first sub-network of the machine learning network, wherein the first sub-network is configured to determine a plurality of contexts of the protected attribute across the identified data samples; determining an impact of each of the plurality of contexts on a second sub-network of the machine learning network, wherein the second sub-network of the machine learning network is configured to classify a given data sample into one of a plurality of classes; and adjusting the second sub-network of the machine learning to account for the impact of at least one of the plurality of contexts on the second sub-network.

Abstract: Method, apparatus, and non-transitory storage medium for neural network based dialogue generation, including receiving an input dialogue context, and generating queries based on the input dialogue context using a query generating neural network. The query generating neural network may be trained using a cheap noisy supervision function. The method may further include retrieving responses from a web-based search engine based on the generated queries, and generating dialogue based on the retrieved responses and the input dialogue context.

Abstract: A method of training a model, an electronic device, and a storage medium are provided, which relate to a field of artificial intelligence, and in particular to a field of reinforcement learning, NLP, etc. The method includes: acquiring a dialogue information; obtaining a predicted information based on the dialogue information by using a first intention recognition model, a first recurrent neural network and a first named entity recognition model; obtaining a machine behavior information based on the predicted information by using a first behavior decision model; acquiring a feedback information for the machine behavior; storing at least one of the predicted information, the machine behavior information, or the feedback information as training data in a database; and performing a model optimization training online based on the training data by using a reinforcement learning algorithm, in response to an amount of the training data reaching a preset data amount.

Abstract: Provided is a method and apparatus for designing and testing an audio codec using quantization based on white noise modeling. A neural network-based audio encoder design method includes generating a quantized latent vector and a reconstructed signal corresponding to an input signal by using a white noise modeling-based quantization process, computing a total loss for training a neural network-based audio codec, based on the input signal, the reconstruction signal, and the quantized latent vector, training the neural network-based audio codec by using the total loss, and validating the trained neural network-based audio codec to select the best neural network-based audio codec.

Abstract: An encoding method, a decoding method, an encoder for performing the encoding method, and a decoder for performing the decoding method are provided. The encoding method includes outputting LP coefficients bitstream and a residual signal by performing an LP analysis on an input signal, outputting a first latent signal obtained by encoding a periodic component of the residual signal, a second latent signal obtained by encoding a non-periodic component of the residual signal, and a weight vector for each of the first latent signal and the second latent signal, using a first neural network module, and outputting a first bitstream obtained by quantizing the first latent signal, a second bitstream obtained by quantizing the second latent signal, and a weight bitstream obtained by quantizing the weight vector, using a quantization module.

Abstract: Implementations relate to an automated assistant that is capable of interacting with non-assistant applications that do not have functionality explicitly provided for interfacing with certain automated assistants. Application data, such as annotation data and/or GUI data, associated with a non-assistant application, can be processed to map such data into an embedding space. An assistant input command can then be processed and mapped to the same embedding space, and a distance from the assistant input command embedding and the non-assistant application data embedding can be determined. When the distance between the assistant input command embedding and the non-assistant application data embedding satisfies threshold(s), the automated assistant can generate instruction(s), for the non-assistant application, that correspond to the non-assistant application data.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for extracting entities from conversation transcript data. One of the methods includes obtaining a conversation transcript sequence, processing the conversation transcript sequence using a span detection neural network configured to generate a set of text token spans; and for each text token span: processing a span representation using an entity name neural network to generate an entity name probability distribution over a set of entity names, each probability in the entity name probability distribution representing a likelihood that a corresponding entity name is a name of the entity referenced by the text token span; and processing the span representation using an entity status neural network to generate an entity status probability distribution over a set of entity statuses.

Abstract: A computing system receives requests from client devices to process voice queries that have been detected in local environments of the client devices. The system identifies that a value that is based on a number of requests to process voice queries received by the system during a specified time interval satisfies one or more criteria. In response, the system triggers analysis of at least some of the requests received during the specified time interval to trigger analysis of at least some received requests to determine a set of requests that each identify a common voice query. The system can generate an electronic fingerprint that indicates a distinctive model of the common voice query. The fingerprint can then be used to detect an illegitimate voice query identified in a request from a client device at a later time.