Abstract: Systems and methods for text input based on neuromuscular information. The system includes a plurality of neuromuscular sensors, arranged on one or more wearable devices, wherein the plurality of neuromuscular sensors is configured to continuously record a plurality of neuromuscular signals from a user, at least one storage device configured to store one or more trained statistical models, and at least one computer processor programmed to obtain the plurality of neuromuscular signals from the plurality of neuromuscular sensors, provide as input to the one or more trained statistical models, the plurality of neuromuscular signals or signals derived from the plurality of neuromuscular signals, and determine based, at least in part, on an output of the one or more trained statistical models, one or more linguistic tokens.

Abstract: The invention relates to a system and method for generating a block of natural language, the system comprising a digital data store capable of storing a data graph according to a data schema, input sub-system for entering natural language data units to the data graph, and a data processor for generating a block of natural language based on the data graph. Further, the data schema allows storage of recursively nested natural language data units and relation data units associated with the natural language data units into the data graph, the relation data units being configured to define relations between natural language data units in the data graph. The data processor is adapted to generate said block of natural language utilizing a plurality of natural language data units and relations between the natural language data units as defined by the relation data units associated therewith.

Abstract: Described herein are systems and methods for augmenting neural speech synthesis networks with low-dimensional trainable speaker embeddings in order to generate speech from different voices from a single model. As a starting point for multi-speaker experiments, improved single-speaker model embodiments, which may be referred to generally as Deep Voice 2 embodiments, were developed, as well as a post-processing neural vocoder for Tacotron (a neural character-to-spectrogram model). New techniques for multi-speaker speech synthesis were performed for both Deep Voice 2 and Tacotron embodiments on two multi-speaker TTS datasets—showing that neural text-to-speech systems can learn hundreds of unique voices from twenty-five minutes of audio per speaker.

Abstract: An approach for optimizing a confidence score threshold that is used to recognize a target word(s) in an audio source. A variety of potential instances of the target word can be detected and classified using an initial confidence score threshold value. Each potential instance of the target word is audibly reviewed and validated by a user. After a determination of the correctness of each potential instance's classification, a different confidence score threshold value can be used to produce an updated set of classification results without requiring the user to revalidate the results. By using a variety of confidence score threshold values to produce various sets of classification results, an optimized confidence threshold setting can be determined for the identified target word based on minimizing errors in the various results. This value can then be applied for future analysis of the target word in an audio source.

Abstract: Improvements in speech recognition in a new domain are provided via the student/teacher training of models for different speech domains. A student model for a new domain is created based on the teacher model trained in an existing domain. The student model is trained in parallel to the operation of the teacher model, with inputs in the new and existing domains respectfully, to develop a neural network that is adapted to recognize speech in the new domain. The data in the new domain may exclude transcription labels but rather are parallelized with the data analyzed in the existing domain analyzed by the teacher model. The outputs from the teacher model are compared with the outputs of the student model and the differences are used to adjust the parameters of the student model to better recognize speech in the second domain.

Abstract: A method includes receiving initial training data associated with a trigger phrase in a device and training a voice model in the device using the initial training data. The voice model is used to identify a plurality of voice commands in the device initiated using the trigger phrase. Collection of additional training data from the plurality of voice commands and retraining of the voice model in the device are iteratively performed using the additional training data. A device includes a microphone and a processor to receive initial training data associated with a trigger phrase using the microphone, train a voice model device using the initial training data, use the voice model to identify a plurality of voice commands initiated using the trigger phrase, and iteratively collect additional training data from the plurality of voice commands and retrain the voice model in the device using the additional training data.

Abstract: Computer-implemented systems and methods are provided for improved generation and control of conversations. A computing device is utilized to control or simulate conversation using estimated contextual cues extracted from profile information or prior responses. The computing device is configured to automatically tailor a flow of a conversation to an effort to improve relevancy and engagement without the need of a human operator to manually tailor the conversation, which, for example, could be impractically expensive. A structured workflow is maintained in the form of a series of conversation decisions, and a machine learning engine is utilized to maintain a continuously trained data structure that generates predictions that bias conversation decisions (e.g., by weighting tree options) for tailoring the conversation flow.

Abstract: A voice processing method for estimating an impression of speech includes: executing an acquisition process that includes acquiring voice signals; executing a feature acquisition process that includes acquiring acoustic features regarding the voice signals from the voice signals; executing a voice-parameter acquisition process that includes acquiring a voice parameter regarding a frame of the voice signals; executing a relative-value determination process that includes determining a relative value between the determined voice parameter and a statistical value of the voice parameter; executing a weight assignment process that includes assigning a weight to the frame of the voice signals in accordance with the relative value; and executing a distribution determination process that includes determining a distribution of the acoustic features, based on the weight assigned to the frame of the voice signals.

Abstract: The present invention provides a computer-aided conversion system and method for generating intelligible speech that uses a transmitter disposed in the nasal cavity of a user and a receiver disposed in pairs with the transmitter, the transmitter transmits a detecting signal in waveform to the nasal cavity of the user, and the receiver receives a reflected wave from the user's nasal cavity. After analyzing the reflected wave, a spectrum corresponding to the acoustic model of an articulatory cavity is obtained. Through the spectrum, the intention in the speaking of the user may be known, that is, the present invention may detect a speech not originated from the vocal cord of the user.

Abstract: An electronic device and a controlling method are provided. The controlling method of the electronic device includes transmitting a signal to a plurality of external devices communicatively connected to the electronic device, receiving, from each of the plurality of external devices, intensity information of the signal sensed by an external device and identification information of an external device, determining at least one external device that is positioned in a same space as the electronic device, from among the plurality of external devices, based on the response signal, designating the at least one external device and the electronic device as a device group, and controlling the device group based on the user command, when a user command is input to at least one device from among the device groups.

Abstract: An system and a method for the detection and reporting of occupational safety incidents are disclosed. The system receives a set of digital records corresponding to reported occupational safety incidents. The system converts each of the digital records from the set of digital records into a common digital format. The system deconstructs the uniform text structure of each digital recorded by a natural language processing module to lemmatize words, remove punctuation, and remove stop words. The system creates a feature vector based on the received deconstructed uniform text structure. The system inputs each feature vector to an ensemble machine learning data model, returning a determination of a possible class or characteristic of occupational safety incident. The system applies a threshold based on a probability to the determination of a possible class. The system submits a subset of the reported occupational safety incidents to a third party system.

Abstract: From a set of information obtained about a user, a profile is constructed representing a speech skill of the user, the set of information including audio, video, and demographic information of the user and other users, the constructing creating new data corresponding to the speech skill of the user in the profile. By correlating analytics of new real-time audio and video information with the new data in the profile, an intervention instruction is triggered automatically, the intervention being directed to change in a voice communication pattern of the user. The intervention is converted to an intervention instruction and the intervention instruction is output in a natural language form. New real-time audio and video information received in response to the spoken natural language instruction is analyzed.

Abstract: Systems and methods are described for determining whether to activate a voice activated device based on a speaking cadence of the user. When the user speaks with a first cadence the system may determine that the user does not intend to activate the device and may accordingly not to trigger a voice activated device. When the user speaks with a second cadence the system may determine that the user does wish to trigger the device and may accordingly trigger the voice activated device.

Abstract: In one implementation, a computer-implemented method includes receiving, at a mobile computing device, ambiguous user input that indicates more than one of a plurality of commands; and determining a current context associated with the mobile computing device that indicates where the mobile computing device is currently located. The method can further include disambiguating the ambiguous user input by selecting a command from the plurality of commands based on the current context associated with the mobile computing device; and causing output associated with performance of the selected command to be provided by the mobile computing device.

Abstract: An electronic device and a controlling method are provided. The controlling method of the electronic device includes transmitting a signal to a plurality of external devices communicatively connected to the electronic device, receiving, from each of the plurality of external devices, intensity information of the signal sensed by an external device and identification information of an external device, determining at least one external device that is positioned in a same space as the electronic device, from among the plurality of external devices, based on the response signal, designating the at least one external device and the electronic device as a device group, and controlling the device group based on the user command, when a user command is input to at least one device from among the device groups.

Abstract: In order to detect a replay attack on a voice biometrics system, a speech signal is received at at least a first microphone and a second microphone. The speech signal has components at first and second frequencies. The method of detection comprises: obtaining information about a position of a source of the first frequency component of the speech signal, relative to the first and second microphones; obtaining information about a position of a source of the second frequency component of the speech signal, relative to the first and second microphones; comparing the position of the source of the first frequency component and the position of the source of the second frequency component; and determining that the speech signal may result from a replay attack if the position of the source of the first frequency component differs from the position of the source of the second frequency component by more than a threshold amount.

Abstract: An information processing apparatus by which smooth communication with a user by voice can be implemented is provided. The information processing apparatus presents a plurality of choices to the user, recognizes utterance contents of the user for selecting one of the plurality of choices, and specifies the choice selected by the user based on whether or not a phrase included in the recognized utterance contents of the user corresponds to a phrase included in a dictionary corresponding to each of the plurality of choices prepared in advance.

Abstract: The present disclosure discloses a method and apparatus for pushing information.

Abstract: Disclosed are systems and methods for improving interactions with and between computers in content searching, generating, hosting and/or providing systems supported by or configured with personal computing devices, servers and/or platforms. The disclosure provides a computerized framework for automatically generating chatbot responses to produce domain-specific responses that mimic native styles unique to particular domains. The disclosed systems and methods construct domain-specific word-graphs based on account activity from specific domains and generate word-patterns. New words obtained from the patterns in the graph are introduced to transform the regular response. The graph is then pruned using data-driven thresholds in order to avoid spurious transformations, and paragraph vectors are also utilized to assign relevance scores to generated patterns such that only the patterns that are contextually similar to the original response (generic/regular response) are used.

Abstract: The present invention is a method and apparatus for narrative content generation using narrative frameworks by receiving a first phrase variation and a second phrase variation and displaying an error indication when the first phrase variation fails to satisfy a criterion relative to the second phrase variation. If there is an error indication, alternate phrase variations are received and compared against the first phrase variation until an alternate phrase variation is selected that has no error indication. Additionally, multiple sets of operators for updating one or more narrative phrases selected for inclusion in the narrative content framework may be utilized to update selected phrases after inclusion in the narrative framework but prior to finalizing the narrative content to be output.