Abstract: A method of speaker identification comprises receiving an audio signal representing speech; performing a first voice biometric process on the audio signal to attempt to identify whether the speech is the speech of an enrolled speaker; and, if the first voice biometric process makes an initial determination that the speech is the speech of an enrolled user, performing a second voice biometric process on the audio signal to attempt to identify whether the speech is the speech of the enrolled speaker. The second voice biometric process is selected to be more discriminative than the first voice biometric process.

Abstract: An apparatus for encoding a speech signal by determining a codebook vector of a speech coding algorithm is provided. The apparatus includes a matrix determiner for determining an autocorrelation matrix R, and a codebook vector determiner for determining the codebook vector depending on the autocorrelation matrix R. The matrix determiner is configured to determine the autocorrelation matrix R by determining vector coefficients of a vector r, wherein the autocorrelation matrix R includes a plurality of rows and a plurality of columns, wherein the vector r indicates one of the columns or one of the rows of the autocorrelation matrix R, wherein R(i, j)=r(|i?j|), wherein R(i, j) indicates the coefficients of the autocorrelation matrix R, wherein i is a first index indicating one of a plurality of rows of the autocorrelation matrix R, and wherein j is a second index indicating one of the plurality of columns of the autocorrelation matrix R.

Abstract: A causality recognizing apparatus includes a candidate vector generating unit configured to receive a causality candidate for generating a candidate vector representing a word sequence forming the candidate; a context vector generating unit generating a context vector representing a context in which noun-phrases of cause and effect parts of the causality candidate appear; a binary pattern vector generating unit, an answer vector generating unit and a related passage vector generating unit, generating a word vector representing background knowledge for determining whether or not there is causality between the noun-phrase included in the cause part and the noun-phrase included in the effect part; and a multicolumn convolutional neural network learned in advance to receive these word vectors and to determine whether or not the causality candidate has causality.

Abstract: An input method includes receiving input end indication information sent by an input module, where the input end indication information indicates that input of a character or a word ends, obtaining a location of a cursor, identifying the input character or word forward from the location of the cursor until a first punctuation input before the character or the word is identified, using the identified character or word as a previous text, and querying a word library for a next text associated with the previous text, and outputting the associated next text to a display module for displaying.

Abstract: Systems and methods are provided for facilitating data object extraction from unstructured documents. Unstructured documents may include data in an unorganized format, such as raw text. The system may use natural language processing to determine characteristics of the terms used in the unstructured document. The system may prompt a user to select terms from the document corresponding in characteristics to properties of a data object being generated. The user may select terms from the document and the system may generate a data object according to the selected terms.

Abstract: Systems and processes for accelerating task performance are provided. An example method includes, at an electronic device with a display, displaying a candidate shortcut affordance associated with a user activity, detecting a first set of inputs corresponding to a selection of the candidate shortcut affordance, in response to detecting the first set of inputs, displaying a first set of candidate task affordances, detecting a second set of inputs corresponding to a selection of a candidate task affordance associated with a first task, displaying a second set of candidate task affordances, detecting a third set of inputs corresponding to a selection of a candidate task affordance associated with a second task, and in response to detecting the second set of inputs and the third set of inputs, associating the first task and the second task with a task sequence for a voice shortcut corresponding to the user activity.

Abstract: A method of speaker identification comprises receiving an audio signal representing speech; performing a first voice biometric process on the audio signal to attempt to identify whether the speech is the speech of an enrolled speaker; and, if the first voice biometric process makes an initial determination that the speech is the speech of an enrolled user, performing a second voice biometric process on the audio signal to attempt to identify whether the speech is the speech of the enrolled speaker. The second voice biometric process is selected to be more discriminative than the first voice biometric process.

Abstract: A device receives document information associated with a document, and receives a request to identify insights in the document information. The device performs, based on the request, natural language processing on the document information to identify words, phrases, and sentences in the document information, and utilizes a first machine learning model with the words, the phrases, and the sentences to identify information indicating abstract insights, concrete insights, and non-insights in the document. The device utilizes a second machine learning model to match the abstract insights with particular concrete insights that are different than the concrete insights, and utilizes a third machine learning model to determine particular insights based on the non-insights. The device generates an insight document that includes the concrete insights, the abstract insights matched with the particular concrete insights, and the particular insights determined based on the non-insights.

Abstract: Artificial agents utilized for voice interactions continue to improve in their capacity to conduct more sophisticated interactions. Rather than just presenting a limited set of options, artificial agents are continuing to narrow the gap between generated speech and natural human speech. A requirement is often in place that spoken interactions be recorded, however, storing speech, even with data compression, is a resource-demanding task. Generated speech may be provided from content, such as text, and speech data. By recording an identifier of the content and associated speech data, storage processing and space requirements can be greatly reduced. Playback may be provided from a waveform of audio provided by the human participant and by selecting the content associated with the content identifier and generating speech of the content utilizing settings provided by the speech data.

Abstract: Aspects of the disclosure describe improving identification of product solutions. An example method includes transcribing in real-time a conversation between a user and an agent into a speech text, processing digital data of the speech text associated with a topic, including parsing the speech text into one or more words and determining collocation information among the one or more words in the speech text. The method also includes providing the one or more words and the collocation information as a first input set to a machine learning engine configured to recommend one or more product solutions from a library of product solutions, generating a recommendation of one or more product solutions for a user based on recommendation parameters for the library of product solutions, and providing the recommendation.

Abstract: In some examples, machine learning based quantification of performance impact of data irregularities may include generating an irregularity feature vector for each text analytics application of a plurality of text analytics applications. Normalized data associated with a corresponding text analytics application may be generated for each text analytics application and based on minimization of irregularities present in un-normalized data associated with the corresponding text analytics application. An un-normalized data machine learning model may be generated for each text analytics application and based on the un-normalized data associated with the corresponding text analytics application. A normalized data machine learning model may be generated for each text analytics application and based on the normalized data associated with the corresponding text analytics application.

Abstract: Exemplary embodiments of restroom monitoring systems having a virtual assistants includes a communications gateway located in a restroom. The communications gateway having a processor, memory, short range communications circuitry, long range communications circuitry, a microphone and a speaker. The communications gateway containing logic for listening for a wake up word and upon detecting a wake up word, capturing a request, logic for processing the request to determine what request is being requested, logic for verifying the request with the requester, and one of a plurality of wave files and a voice synthesizer. The system further includes one or more dispensers located in the restroom. The one or more dispensers having short range communications circuitry for communicating status or product level to the communications gateway.

Abstract: A method of training an obfuscation network for obfuscating original data to protect personal information is provided. The method includes steps of a learning device, (a) inputting acquired training data into an obfuscation network to obfuscate the training data and inputting the obfuscated training data into an augmentation network to augment the obfuscated training data; (b) (i) inputting the augmented obfuscated training data into a learning network to generate first characteristic information and (ii) inputting the training data into the learning network to generate second characteristic information; and (c) training the obfuscation network such that (i) a first error, calculated by using the first and the second characteristic information, is minimized and (ii) a second error, calculated by using (ii-1) modified training data or modified obfuscated training data, and (ii-2) the obfuscated training data or the augmented obfuscated training data, is maximized.

Abstract: Systems and methods for contextual search query revision are disclosed. A user utterance including at least one semantic component is received and a plurality of candidate n-grams including the at least one semantic component and at least one additional semantic component selected from a set of prior semantic components is generated. A probability that each of the plurality of candidate n-grams is an intended n-gram is calculated and a selected one of the plurality of candidate n-grams is output based on the probability.

Abstract: The present disclosure relates to an apparatus which communicates with a voice recognition device, and a method for controlling an apparatus with a voice recognition capability which operates in the Internet of Things environment configured by a 5G communication network. According to an exemplary embodiment of the present disclosure, an apparatus with a voice recognition capability includes a container which has one open surface and accommodates objects therein, a door which opens/closes the container, a sensor which senses an open/closed state of the door, a microphone which receives an external voice, a voice recognizer which recognizes a voice command received from the microphone, and a controller which controls an active state and an inactive state of the voice recognizer, in which the controller may predict whether the voice recognizer needs to be activated using a deep neural network model learned through the machine learning.

Abstract: A technique for encoding a multichannel audio encoding is provided that includes quantizing a set of first LP filter coefficients for an audio signal in a first channel using a predefined first quantizer; and quantizing a set of second LP filter coefficients for an audio signal in a second channel on the basis of the quantized set of first LP filter coefficients. The quantization of the set of second LP filter coefficients includes: deriving, on basis of the quantized set of first LP filter coefficients by using a predefined predictor, a set of predicted LP filter coefficients for the audio signal in said second channel, computing prediction error as a difference between respective LP coefficients of the set of second LP filter coefficients and the set of predicted LP filter coefficients, and quantizing the prediction error.

Abstract: Methods, systems, apparatus, including computer programs encoded on a computer storage medium, for a user device to learn offline voice actions. In one aspect, the method includes actions of detecting, by the user device, an utterance at a first time when the user device is connected to a server by a network, providing, by the user device, the utterance to the server using the network, receiving, by the user device and from the server, an update to the grammar of the user device, detecting, by the user device, a subsequent utterance of the utterance at a second time when the user device is not connected to the server by a network, and in response to detecting, by the user device, the subsequent utterance of the utterance at the second time, identifying, by the user device, an operation to perform based on (i) the subsequent utterance, and (ii) the updated grammar.

Abstract: A method of determining a distribution of bits for coding a transition frame, said method being implemented in a coder/decoder for coding/decoding a digital signal, the transition frame being preceded by a predictive coded preceding frame, coding the transition frame comprising transform coding and predictive coding a single sub-frame of the transition frame, the method comprising the following steps: assigning a bit rate for predictive coding the transition sub-frame, said bit rate being equal to the minimum between the bit rate for transform coding the transition frame and a first predetermined bit rate value; determining a first number of bits allocated for predictive coding the transition sub-frame for said bit rate; and calculating a second number of bits allocated for transform coding the transition frame from the first number of bits and a number of bits available for coding the transition frame.

Abstract: A method and apparatus include receiving a text input that includes a sequence of text components. Respective temporal durations of the text components are determined using a duration model. A spectrogram frame is generated based on the duration model. An audio waveform is generated based on the spectrogram frame. Video information is generated based on the audio waveform. The audio waveform is provided as an output along with a corresponding video.

Abstract: In general, techniques are described for compressing decomposed representations of a sound field. A device comprising one or more processors may be configured to perform the techniques. The one or more processors may be configured to obtain a bitstream comprising a compressed version of a spatial component of a sound field, the spatial component generated by performing a vector based synthesis with respect to a plurality of spherical harmonic coefficients.