Abstract: A method and an apparatus for packet loss concealment, and a decoding method and an apparatus employing same are provided. A method for time domain packet loss concealment includes checking whether a current frame is either an erased frame or a good frame after the erased frame, when the current frame is either the erased frame or the good frame after the erased frame, obtaining signal characteristics, selecting one of a phase matching tool and a smoothing tool based on a plurality of parameters including the signal characteristics, and performing a packet loss concealment processing on the current frame based on the selected tool.

Abstract: An encoding apparatus performing encoding by an encoding process in which bits are preferentially assigned to a low side to obtain a spectrum code, the encoding apparatus judging whether a sound signal is a hissing sound or not, obtaining and encoding, if the encoding apparatus judges that the sound signal is a hissing sound, what is obtained by exchanging all or a part of a spectrum existing on a lower side than a predetermined frequency in a frequency spectrum sequence of the sound signal for all or a part of a spectrum existing on a higher side of the predetermined frequency in the frequency spectrum sequence, the number of all or the part of the high-side frequency spectrum sequence being the same as the number of all or the part of the low-side frequency spectrum sequence.

Abstract: Mechanisms are provided to implement an annotation mechanism allows users to annotate documents with annotations for processing by a cognitive medical system. The annotation mechanism receives, via a user interface, a user selection of an electronic document for annotation, and determines one or more domains associated with the selected electronic document from an analysis of metadata associated with the selected electronic document. The annotation mechanism retrieves a predefined set of annotations associated with each determined domain, and presents the predefined set of annotations as user selectable elements. The annotation mechanism receives, via the user interface, a selection of one or more annotations in the predefined set of annotations to be associated with the selected portion of the selected electronic document, and generates annotation metadata associating the selected portion using the selected one or more annotations.

Abstract: Embodiments of the present disclosure provide a language conversion method and apparatus based on syntactic linearity and a non-transitory computer-readable storage medium. The method includes: encoding a source sentence to be converted by using a preset encoder to determine a first vector and a second vector corresponding to the source sentence; determining a current mask vector according to a preset rule, in which the mask vector is configured to modify vectors output by the preset encoder; determining a third vector according to target language characters corresponding to source characters located before a first source character; and decoding the first vector, the second vector, the mask vector, and the third vector by using a preset decoder to generate a target character corresponding to the first source character.

Abstract: A computer-implemented method includes receiving, by a natural language processing (NLP) annotator, an input text that is to be annotated. The method further includes determining, by the NLP annotator, a user setting that indicates an aggressiveness level of annotation to be used to annotate the input text. The method further includes selecting, by the NLP annotator, from a plurality of dictionaries, a first dictionary based at least in part on the aggressiveness level. The method further includes generating, by the NLP annotator, annotated text of the input text based at least in part on the first dictionary. The method further includes outputting, by the NLP annotator, the annotated text.

Abstract: An apparatus for generating an error concealment signal includes an LPC (linear prediction coding) representation generator for generating a first replacement LPC representation and a different second replacement LPC representation; an LPC synthesizer for filtering a first codebook information using the first replacement representation to obtain a first replacement signal and for filtering a different second codebook information using the second replacement LPC representation to obtain a second replacement signal; and a replacement signal combiner for combining the first replacement signal and the second replacement signal to obtain the error concealment signal.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for speech recognition using complex linear projection are disclosed. In one aspect, a method includes the actions of receiving audio data corresponding to an utterance. The method further includes generating frequency domain data using the audio data. The method further includes processing the frequency domain data using complex linear projection. The method further includes providing the processed frequency domain data to a neural network trained as an acoustic model. The method further includes generating a transcription for the utterance that is determined based at least on output that the neural network provides in response to receiving the processed frequency domain data.

Abstract: In apparatus, methods, and programs for selecting pitch lag, an encoder obtains a first and a second estimates of a pitch lag for a current frame. A selected value is chosen by selection between the first and the second estimates, based on a first and a second correlation measurements. The second estimate is conditioned by the pitch lag selected at the previous frame. The selection is based on a comparison between: a downscaled version of a first correlation measurement associated to the current frame and obtained at a lag corresponding to the first estimate; and a second correlation measurement associated to the current frame and obtained at a lag corresponding to the second estimate.

Abstract: A method for performing voice analysis includes storing, in a database, a simulation file for conducting a training session with a user, the simulation file including at least a script, storing desired attributes associated with the simulation file, retrieving the simulation file from the database and providing a user interface to conduct the voice analysis using the simulation file from the database, receiving one or more voice impressions from a user and analyzing, at an audio analysis tool, at least one of the voice impressions of the user determining, at the audio analysis tool, attributes of the at least one voice impression in response to analyzing the at least one voice impression and comparing, at the audio analysis tool, the determined attributes to the desired attributes associated with the simulation file. The method provides, by the client application, feedback to the user based on the comparison.

Abstract: Provided is an artificial intelligence (AI) answering system including a user question receiver configured to receive a user question from a user terminal; a first question extender configured to generate a question template by analyzing the user question and determine whether the user question and the generated question template match; a second question extender configured to generate a similar question template by using a natural language processing and a deep learning model when the user question and the generated question template do not match; a training data builder configured to generate training data for training the second question extender by using an neural machine translation (NMT) engine; and a question answering unit configured to transmit a user question result derived through the first question extender or the second question extender to the user terminal.

Abstract: In present disclosure, a data processing method, a data processing device, and an apparatus for data processing are provided. The method specifically includes: receiving a source language speech input by a target user; determining, based on the source language speech, a target acoustic model from a preset acoustic model library, the acoustic model library including at least two acoustic models corresponding to different timbre characteristics; converting, based on the target acoustic model, the source language speech into a target language speech; and outputting the target language speech. According to the embodiments of the present disclosure, the recognition degree of the speaker corresponding to the target language speech output by the translation device can be increased, and the effect of user communication can be improved.

Abstract: An audio coding method, comprising: obtaining an ith audio frame in n consecutive audio frames and obtaining an ith piece of coded data and an ith piece of redundant data based on the ith audio frame, wherein the ith piece of coded data is obtained by coding the ith audio frame, and the ith piece of redundant data is obtained by coding and buffering the ith audio frame, wherein n is a positive integer, and 1?i?n; and packing the ith piece of coded data and at most m pieces of redundant data before the ith piece of redundant data into an ith audio data packet, wherein m is a preset positive integer. An audio decoding method and a computer device are further provided.

Abstract: Performance of Automatic Speech Recognition (ASR) for robustness against real world noises and channel distortions is critical. Embodiments herein provide method and system for generating synthetic multi-conditioned data sets for additive noise and channel distortion for training multi-conditioned acoustic models for robust ASR. The method provides a generative noise model generating plurality of types of noise signals for additive noise based on weighted linear combination of plurality of noise basis signals and channel distortion based on estimated channel responses. The generative noise model is a parametric model, wherein basis function selection, number of basis functions to be combined linearly and weightages to be applied to the combinations is tunable, thereby enabling generation of wide variety of noise signals. Further, the noise signals are added to set of training speech utterances under set of constraints providing the multi-conditioned data sets, imitating real world effects.

Abstract: A system to convert sequences of human thought representations into coherent stories, in association with a language understanding system is disclosed. Said system comprises: an entity dereferencing and enrichment module, an anomaly detecting unit that comprises: a context anomaly module, and a meaning anomaly and reinforcement module; an inter thought representation reasoning and transformation unit; an entity knowledge base; a thought representation knowledge base; and an output thought representation cloud. The system takes sequences of thought representations as input and tries to make sense out of them. The system is used in association with any type of language understanding system for creating meaning out of the sequence of thoughts.

Abstract: A computer-implemented method is provided for transferring a target text style using Reinforcement Learning (RL). The method includes pre-determining, by a Long Short-Term Memory (LSTM) Neural Network (NN), the target text style of a target-style natural language sentence. The method further includes transforming, by a hardware processor using the LSTM NN, a source-style natural language sentence into the target-style natural language sentence that maintains the target text style of the target-style natural language sentence. The method also includes calculating an accuracy rating of a transformation of the source-style natural language sentence into the target-style natural language sentence based upon rewards relating to at least the target text style of the source-style natural language sentence.

Abstract: The present disclosure provides a method, a browser client, and a server for reading web page information by speech. The browser client is installed with a text to speech (TTS) engine. The method includes: sending, by a browser client, a page access request to a server, where the page access request includes a page address and TTS identity information; receiving, by the browser client, response data returned by the server, where the response data includes a TTS standard version number determined by the server according to the TTS identity information, and TTS page data corresponding to the page address; and reading, by the browser client, the TTS page data by speech according to the TTS standard version number by using a TTS engine. In the present disclosure, page information is read by speech by using the TTS engine installed on the browser client.

Abstract: A processing system detects a period of non-voice activity and compares its duration to a cutoff period. The system adapts the cutoff period based on parsing previously-recognized speech to determine, according to a model, such as a machine-learned model, the probability that the speech recognized so far is a prefix to a longer complete utterance. The cutoff period is longer when a parse of previously recognized speech has a high probability of being a prefix of a longer utterance.

Abstract: Techniques to be used in natural language understanding (NLU) are described. For example, a NLU service to receive a request to analyze a written or spoken utterance; tokenize the received utterance; generate one or more labels corresponding to a substring of the tokenized received utterance, each of the labels including one or more slot types, by: for each path of a grammar-based finite state transducer (FST) data structure that includes instructions, traversing the path as far as possible for matches from a previous breakpoint, while maintaining i) locations of branching points and snapshots at those branching points and ii) an indication of which paths have been traversed, and recording a result of each path traversal as a generated label; resolve the one or more generated labels into machine-readable values; and output a result is described.

Abstract: Systems, apparatus, methods, and articles of manufacture for cooperatively-overlapped and Artificial Intelligence (AI)-managed interfaces. For example, multiple cooperatively and/or partially overlapped interfaces may be provided (e.g., via an electronic and/or touch-screen device), with such interfaces being dynamically managed by various AI components, such as natural language processing, machine learning techniques, and/or neural network data processing.

Abstract: A system, method and computer-readable storage device are disclosed signing a voicemail and confirming an identity of the speaker. A method includes receiving a request to verify a speaker associated with a communication to a recipient, receiving first data from the speaker in connection with the communication, accessing second data associated with the speaker to verify the speaker, determining whether a match exists between the first data and the second data to yield a determination, retrieving a communication address of the recipient, generating a notification for the recipient, wherein the notification reports on the determination and transmitting the notification to the recipient at the communication address.