Abstract: A method for reducing response time in a speech interface including constructing a partially completed word sequence from a partially received utterance from a speaker received by an audio sensor, modeling a remainder portion using a processor based on a rich predictive model to predict the remainder portion, and responding to the partially completed word sequence and the predicted remainder portion using a natural language vocalization generator with a vocalization, wherein the vocalization is prepared before a complete utterance is received from the speaker and conveyed to the speaker by an audio transducer.

Abstract: Disclosed are a system and a method of providing customized content by using a sound, the system including: multiple smart devices outputting content for each channel received from a content provider; multiple user terminals configured to: receive a user's voice for a preset text for voice recognition to register the user's voice as voice data for user identification, and receive waveform data of a sound output from the smart device and voice data of the user for transmission when a voice corresponding to the registered voice data for user identification is recognized while the smart device is in operation; and a recommendation server configured to: collect the waveform data for the content that is possibly output from the smart device, and analyze the voice data and the waveform data to generate interest information of the user when the waveform data and the voice data are transmitted from the user terminal.

Abstract: The present disclosure provides a method and an apparatus for semantic recognition, and a system for human-machine dialog. In the method, a Pinyin sequence of a sentence to be recognized is obtained. The Pinyin sequence includes a plurality of Pinyin segments. Then, word vectors of the plurality of Pinyin segments are obtained. Next, the word vectors of the plurality of Pinyin segments are combined into a sentence vector of the sentence to be recognized. Based on the sentence vector of the sentence to be recognized, an output vector of the sentence to be recognized is obtained by using a neural network. Based on the output vector of the sentence to be recognized, a reference sentence semantically similar to the sentence to be recognized is determined. Then, a semantic meaning of the sentence to be recognized is recognized as a semantic meaning of the reference sentence.

Abstract: A method and apparatus of a device that builds a target using a plurality of processing units is described. In an exemplary embodiment, the device receives an input audio track having a first duration, the input audio track having a plurality of points. The device further generates a transition matrix of the input audio track, wherein the transition matrix indicates a similarity metric between different pairs of the plurality of points. In addition, the device determines a set of jump points using the different pairs of the plurality of points. The device additionally generates the rearranged audio track using the set of jump points, wherein the rearranged audio track has second duration and the second duration is different than the first duration.

Abstract: Embodiments describe a method for speech endpoint detection including receiving identification data for a first state associated with a first frame of speech data from a WFST language model, determining that the first frame of the speech data includes silence data, incrementing a silence counter associated with the first state, copying a value of the silence counter of the first state to a corresponding silence counter field in a second state associated with the first state in an active state list, and determining that the value of the silence counter for the first state is above a silence threshold. The method further includes, determining that an endpoint of the speech has occurred in response to determining that the silence counter is above the silence threshold, and outputting text data representing a plurality of words determined from the speech data that was received prior to the endpoint.

Abstract: The invention relates to a method for enhanced voiceprint authentication. The method includes receiving an utterance from a user, and determining that a portion of the utterance matches a pre-determined keyword. Also, the method includes authenticating the user by comparing the portion of the utterance with a voiceprint that is associated with the pre-determined keyword. Further, the method includes identifying a resource associated with the pre-determined keyword while comparing the portion of the utterance with the voiceprint. Still yet, the method includes accessing the resource in response to authenticating the user based on the comparison.

Abstract: Natural language content can be provided by multiple and various sources. Once in text format, such content may be provided to various systems for further processing to identify one or more propositions. The relationship between each proposition may be identified and ordered according to the identified relationships. A visual display may be generated to illustrate the identified propositions and relationship, as well as identify any propositions that may be missing, unsupported, or other characteristic thereof.

Abstract: Methods, systems, and apparatus for receiving audio data corresponding to a user utterance and context data, identifying an initial set of one or more n-grams from the context data, generating an expanded set of one or more n-grams based on the initial set of n-grams, adjusting a language model based at least on the expanded set of n-grams, determining one or more speech recognition candidates for at least a portion of the user utterance using the adjusted language model, adjusting a score for a particular speech recognition candidate determined to be included in the expanded set of n-grams, determining a transcription of user utterance that includes at least one of the one or more speech recognition candidates, and providing the transcription of the user utterance for output.

Abstract: The present disclosure provides a method, an apparatus and a storage medium for a wake-up processing of an application, a first voice data input by a user is collected and recognized, and a wake-up operation is performed on a target application each time when it is recognized that a wake-up word of the target application is included in the first voice data, where the wake-up word of the target application appears one or more times in the first voice data. The method, apparatus and storage medium for the wake-up processing of the application provided by the present disclosure can wake up the target application when the wake-up word appears one or more times in the first voice data input by the user, thereby improving a wake-up accuracy of the application.

Abstract: A speech/audio bitstream decoding method includes acquiring a speech/audio decoding parameter of a current speech/audio frame, where the foregoing current speech/audio frame is a redundant decoded frame or a speech/audio frame previous to the foregoing current speech/audio frame is a redundant decoded frame, performing post processing on the acquired speech/audio decoding parameter according to speech/audio parameters of X speech/audio frames, where the foregoing X speech/audio frames include M speech/audio frames previous to the foregoing current speech/audio frame and/or N speech/audio frames next to the foregoing current speech/audio frame, and recovering a speech/audio signal using the post-processed speech/audio decoding parameter of the foregoing current speech/audio frame. The technical solutions of the speech/audio bitstream decoding method help improve quality of an output speech/audio signal.

Abstract: A document analysis system trains a document embedding model configured to receive a set of word embeddings for an ordered set of words in a document and generate a document embedding for the document. The document embedding is a representation of the document in a latent space that characterizes the document with respect to properties such as structure, content, and sentiment. The document embedding may represent a prediction of a set of words that follow the last word in the ordered set of words of the document. The document embedding model may be associated with a convolutional neural network (CNN) architecture that includes one or more convolutional layers. The CNN architecture of the document embedding model allows the document analysis system to overcome various difficulties of existing document embedding models, and allows the document analysis system to easily process variable-length documents that include a variable number of words.

Abstract: Systems and methods for providing customized output based on a user preference in a distributed system are provided. In example embodiments, a meeting server or system receives audio streams from a plurality of distributed devices involved in an intelligent meeting. The meeting system identifies a user corresponding to a distributed device of the plurality of distributed devices and determines a preferred language of the user. A transcript from the received audio streams is generated. The meeting system translates the transcript into the preferred language of the user to form a translated transcript. The translated transcript is provided to the distributed device of the user.

Abstract: Disclosed herein are example embodiments of an improved narrative generation system where an analysis service that executes data analysis logic that supports story generation is segregated from an authoring service that executes authoring logic for story generation through an interface. The analysis service may comprise a plurality of analysis applications and a plurality of analysis libraries, where the analysis applications can be segregated from the analysis libraries through another interface. Accordingly, when the authoring service needs analysis from the analysis service, the authoring service can invoke the analysis service through the interface; and when an analysis application needs analysis from an analysis library, the analysis application can invoke the analysis library through the another interface.

Abstract: Vector Quantizer and method therein for vector quantization, e.g. in a transform audio codec. The method comprises comparing an input target vector with four centroids C0, C1, C0,flip and C1,flip, wherein centroid C0,flip is a flipped version of centroid C0 and centroid C1,flip is a flipped version of centroid C1, each centroid representing a respective class of codevectors. A starting point for a search related to the input target vector in the codebook is determined, based on the comparison. A search is performed in the codebook, starting at the determined starting point, and a codevector is identified to represent the input target vector. A number of input target vectors per block or time segment is variable. A search space is dynamically adjusted to the number of input target vectors. The codevectors are sorted according to a distortion measure reflecting the distance between each codevector and the centroids C0 and C1.

Abstract: A system for artificial intelligence (AI) based omnichannel data communications is disclosed. The system may comprise a data store to store and manage data within a network, a server to facilitate operations using information from the one or more data stores, and an analytics subsystem to communicate with the server and the data store in the network. The analytics subsystem may comprise a data access interface to receive communication data from a user via a first communication channel. The analytics subsystem may comprise a processor to provide a proactive, adaptive, and intelligent conversation with the user via a second communication channel. The system may also comprise an output interface to transmit the at least one appropriate response to the user at a user device using any number of AI-based and data processing techniques to better identify customer touchpoints and improve customer journey and experience during customer engagement scenarios.

Abstract: The present invention provides a speech processing method, which includes: receiving speech information; recognizing the speech information to convert the speech information into a plurality of pieces of text information; displaying at least one pieces of the text information; and receiving a selection signal to respond according to selected text information. The present invention further provides a speech processing device. According to the speech processing method and device provided by embodiments of the present invention, a user can select and confirm by clicking on displayed text information after the speech information is recognized, so that the selected text information is guaranteed to be coincident with a received speech information, thus further increasing accuracy and convenience of the speech processing and providing the user with a good experience effect.

Abstract: Disclosed herein are example embodiments of an improved narrative generation system where an analysis service that executes data analysis logic, including data re-organization logic, that supports story generation is segregated from an authoring service that executes authoring logic for story generation through an interface. Accordingly, when the authoring service needs analysis from the analysis service, it can invoke the analysis service through the interface. By exposing the analysis service to the authoring service through the shared interface, the details of the logic underlying the analysis service are shielded from the authoring service (and vice versa where the details of the authoring service are shielded from the analysis service). Through parameterization of operating variables, the analysis service can thus be designed as a generalized data analysis service that can operate in a number of different content verticals with respect to a variety of different story types.

Abstract: Described herein is a mechanism for improving the accuracy of a language model interpreting short input utterances. A language model operates in a stateless manner, only ascertaining the intents and/or entities associated with a presented input utterance. To increase the accuracy, two language understanding models are trained. One is trained using only input utterances. The second is trained using input utterance-prior dialog context pairs. The prior dialog context is previous intents and/or entities already determined from the utterances in prior turns of the dialog. When input is received, the language understanding model decides whether the input comprises only an utterance or an utterance and prior dialog context. The appropriate trained machine learning model is selected and the intents and/or entities associated with the input determined by the selected machine learning model.

Abstract: A response system which realizes an automatic chatting response while reducing investment to a computer necessary for a learning process in artificial intelligence is provided. A chat pattern generation unit of a terminal device generates information on a chat pattern which is a pattern of input fragments which are inputs in a chat and response fragments which are responses to the input fragments. A learning processing unit performs a learning process on the relationship between input text corresponding to the input fragments and the response fragments on the basis of the generated chat pattern and outputs input interpretation knowledge data acquired as a result of the learning process. A response knowledge data transmission unit transmits response knowledge data based on response fragments included in the chat pattern generated by the chat pattern generation unit to an automatic response server device.

Abstract: The present teaching relates to method, system, medium, and implementations for configuring 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. One or more preferences of the user are identified from the obtained information and are used to select, from the plurality of selectable profiles, a first selected profile, which specifies parameters to be used to control a manner by which the animatronic device is to communicate with the user. The animatronic device is then configured based on the first selected profile for carrying out the dialogue in the manner specified.