Abstract: This document relates to machine learning. One example includes a method or technique that can be performed on a computing device. The method or technique can include obtaining a task-adapted generative model that has been tuned using one or more task-specific seed examples. The method or technique can also include inputting dialog acts into the task-adapted generative model and obtaining synthetic utterances that are output by the task-adapted generative model. The method or technique can also include populating a synthetic training corpus with synthetic training examples that include the synthetic utterances. The synthetic training corpus may be suitable for training a natural language understanding model.

Abstract: An extended role play-based utterance set generation apparatus includes a first data store storing role play-based utterance sets and a second data store storing non-role-played utterance sets. The role play-based utterance sets include a first query and a role play-based response to the query. The non-role-played utterance sets include a second query and a non-role-played response to the query. The disclosed technology determines similarity between the role play-based response and the non-role-played response. Upon determining that the role play-based response is the same or similar to the non-role-played response, the disclosed technology generates an association between the role play-based response and the second query and extends the role play-based utterance sets in the first data store with the second query.

Abstract: An apparatus including a user input receiver; a user voice input receiver; a display; and a processor. The processor is configured to: (a) based on a user input being received through the user input receiver, perform a function corresponding to voice input state for receiving a user voice input; (b) receive a user voice input through the user voice input receiver; (c) identify whether or not a text corresponding to the received user voice input is related to a pre-registered voice command or a prohibited expression; and (d) based on the text being related to the pre-registered voice command or the prohibited expression, control the display to display an indicator that the text is related to the pre-registered voice command or the prohibited expression. A method and non-transitory computer-readable medium are also provided.

Abstract: Memory deterioration detection and evaluation includes capturing human utterances with a voice interface and generating, for a user, a human utterances corpus that comprises human utterances selected from the plurality of human utterances based on meanings of the human utterances as determined by natural language processing by a computer processor. Based on data generated in response to signals sensed by one or more sensing devices operatively coupled with the computer processor, contextual information corresponding to one or more human utterances of the corpus is determined. Patterns among the corpus of human utterances are recognized based on pattern recognition performed by the computer processor using one or more machine learning models. Based on the pattern recognition a change in memory functioning of the user is identified. The identified change is classified, based on the contextual information, as to whether the change is likely due to memory impairment of the user.

Abstract: Methods, systems, and computer program products for image object disambiguation resolution are provided herein. An example of a method includes: obtaining a group of classification labels and corresponding confidence values for an object in an image; using a wordweb to determine one or more properties that distinguish between at least a first one of the classification labels and at least a second one of the classification labels within the group; selecting a first property from the properties to generate a question based on information indicating a level of prior knowledge of the user with each of the properties and each of the one or more labels; assigning a belief score to an answer; and determining whether to present at least a second question to verify the first answer based on a comparison of the belief score to a belief threshold value.

Abstract: A method may include obtaining first features of first audio data that includes speech and obtaining second features of second audio data that is a revoicing of the first audio data. The method may further include providing the first features and the second features to an automatic speech recognition system and obtaining a single transcription generated by the automatic speech recognition system using the first features and the second features.

Abstract: Disclosed is a server for supporting a communication environment between different electronic devices. The server includes a communication circuit, a memory, and a processor. The processor is electrically connected to the communication circuit and the memory. The processor is configured to receive a first voice signal transmitted from a second electronic device to a first electronic device through the communication circuit. The Processor is also configured to allow the first electronic device to transmit network connection information for connecting with the server to the second electronic device based on whether the first voice signal corresponds to a second voice signal stored in the memory.

Abstract: A method for providing natural language conversation is implemented by an interactive agent system. The method for providing natural language conversation, according to an embodiment of the present invention includes receiving a natural language input; determining a user intent based on the natural language input by processing the natural language input, and providing a natural language response corresponding to the natural language input, based on at least one of the natural language input and the determined user intent. The natural language response may be provided by determining whether a predetermined first condition is satisfied, providing a natural language response belonging to a category of substantial replies when the first condition is satisfied, determining whether a predetermined second condition is satisfied when the first condition is not satisfied, and providing a natural language response belonging to a category of interjections when the second condition is satisfied.

Abstract: A semantic recognition method and a semantic recognition device are provided. A spectrogram of a speech signal is generated. At least one keyword of the spectrogram is detected by inputting the spectrogram into a neural network model. A semantic category to which each of the at least one keyword belongs is distinguished. A semantic intention of the speech signal is determined according to the at least one keyword and the semantic category of the at least one keyword.

Abstract: Included are: an apparatus function information acquiring unit for acquiring apparatus function information in which a target apparatus and one or more target functions to be executed by the target apparatus, which are determined on the basis of uttered speech, are associated with each other; a procedure determining unit for determining one or more manual operations for executing the one or more target functions and an order of the one or more manual operations on the basis of the apparatus function information acquired by the apparatus function information acquiring unit; and an operation command transmission controlling unit for sequentially transmitting, to the target apparatus, operation commands for outputting operation response output control information corresponding to each of the one or more manual operations in accordance with the order of the one or more manual operations determined by the procedure determining unit.

Abstract: According to one embodiment of the present invention, a server comprises at least one communication interface, at least one processor operatively connected to the communication interface, and at least one memory operatively connected to the processor, wherein the memory store instructions configured to, when executed, cause the processor: receives, from a first electronic device, first input voice data including a first request for conducting a first task by using a second electronic device by user's utterance; determines or receives a state of the first electronic device; and provides a first external electronic device with a first response related to control of the state of the first electronic device. Various other embodiments are possible.

Abstract: A method and a device for increasing the perception of acoustic events, particularly for increasing musical sensitivity. As high a correlation as possible between heard and felt perceptions can be achieved by the conversion of the musical signal into vibrations on the skin, the local impact distribution of the filtered musical signals, the emphasis of the dominant musical signals by expanding the extent of the impact, the transfer of the signal portions in the non-feelable range into the feelable range, and the variable base spectrum adapting to the current musical spectrum.

Abstract: A contextual hashtag generation method, system, and computer program product include receiving content from an online source, identifying a set of contextual indicators for the content, determining an entity-desired outcome for the content, and generating a hashtag for the content using the set of contextual indicators while maximizing the entity-desired outcome.

Abstract: The application provides a virtual object image display method and apparatus, an electronic device and a storage medium, relates to the field of artificial intelligence, in particular to the field of computer vision and deep learning, and may be applied to virtual object dialogue scenarios. The specific implementation scheme includes: segmenting acquired voice to obtain voice segments; predicting lip shape sequence information for the voice segments; searching for a corresponding lip shape image sequence based on the lip shape sequence information; performing lip fusion between the lip shape image sequence and a virtual object baseplate to obtain a virtual object image; displaying the virtual object image. The application improves ability to obtain virtual object image.

Abstract: A computer-implemented method of false keyphrase rejection comprises receiving a captured audio signal of human speech including one or more keyphrases that trigger an action. It also comprises detecting whether or not at least part of the speech is spoken by at least one computer originated voice. The method also has an operation of omitting the triggering of the action at least partly due to the computer originated voice being recognized in the speech.

Abstract: A system for authenticating digital contents includes a computing platform having a hardware processor and a memory storing a software code. According to one implementation, the hardware processor executes the software code to receive digital content, identify an image of a person depicted in the digital content, determine an ear shape parameter of the person depicted in the image, determine another biometric parameter of the person depicted in the image, and calculate a ratio of the ear shape parameter of the person depicted in the image to the biometric parameter of the person depicted in the image. The hardware processor is also configured to execute the software code to perform a comparison of the calculated ratio with a predetermined value, and determine whether the person depicted in the image is an authentic depiction of the person based on the comparison of the calculated ratio with the predetermined value.

Abstract: A speech synthesis model training device includes one or more hardware processors configured to perform the following. Storing, in a speech corpus storing unit, speech data, and pitch mark information and context information of the speech data. From the speech data, analyzing acoustic feature parameters at each pitch mark timing in pitch mark information. From the acoustic feature parameters analyzed, training a statistical model which has a plurality of states and which includes an output distribution of acoustic feature parameters including pitch feature parameters and a duration distribution based on timing parameters.

Abstract: Techniques herein relate to improving quality of classification models for differentiating different user intents by improving the quality of training samples used to train the classification models. Pairs of user intents that are difficult to differentiate by classification models trained using the given training samples are identified based upon distinguishability scores (e.g., F-scores). For each of the identified pairs of intents, pairs of training samples each including a training sample associated with a first intent and a training sample associated with a second intent in the pair of intents are ranked based upon a similarity score between the two training samples in each pair of training samples. A particular pair of training samples with a highest similarity score is selected and provided as output with a suggestion for modifying the particular pair of training samples.

Abstract: Embodiments relate to an audio processing unit that includes a bitstream payload deformatter and a decoding subsystem. The decoding subsystem is coupled to the bitstream payload deformatter and configured to decode at least a portion of a block of an encoded audio bitstream. The block includes a fill element with an identifier indicating a start of the fill element and fill data after the identifier. The fill data includes at least one flag identifying whether a base form of spectral band replication or an enhanced form of spectral band replication is to be performed on audio content of the block. The identifier is a three bit unsigned integer transmitted most significant bit first and having a value of 0x6.

Abstract: A technique for constructing a model supporting a plurality of domains is disclosed. In the technique, a plurality of teacher models, each of which is specialized for different one of the plurality of the domains, is prepared. A plurality of training data collections, each of which is collected for different one of the plurality of the domains, is obtained. A plurality of soft label sets is generated by inputting each training data in the plurality of the training data collections into corresponding one of the plurality of the teacher models. A student model is trained using the plurality of the soft label sets.