Abstract: Embodiments of the present invention provide an approach for estimating the accuracy of a transcription of a voice recording. Specifically, in a typical embodiment, each word of a transcription of a voice recording is checked against a customer-specific dictionary and/or a common language dictionary. The number of words not found in either dictionary is determined. An accuracy number for the transcription is calculated from the number of said words not found and the total number of words in the transcription.

Abstract: One embodiment of the present invention provides a system for facilitating information in multiple languages based on an optical code. During operation, the system scans an optical code accompanying a text phrase and receives a location of a conversion server embedded in the optical code. The system then determines one or more target languages for obtaining the text phrase in the target languages and sends a query message to the conversion server based on the retrieved location. The query message comprises a list of the one or more target languages.

Abstract: Methods are presented for generating a natural language model. The method may comprise: ingesting training data representative of documents to be analyzed by the natural language model, generating a hierarchical data structure comprising at least two topical nodes within which the training data is to be subdivided into by the natural language model, selecting a plurality of documents among the training data to be annotated, generating an annotation prompt for each document configured to elicit an annotation about said document indicating which node among the at least two topical nodes said document is to be classified into, receiving the annotation based on the annotation prompt; and generating the natural language model using an adaptive machine learning process configured to determine patterns among the annotations for how the documents in the training data are to be subdivided according to the at least two topical nodes of the hierarchical data structure.

Abstract: Provided herein is an electronic device and method of voice recognition, the method including analyzing an audio signal of a first frame when the audio signal is input and extracting a first feature value; determining a similarity between the first feature value extracted from the audio signal of the first frame and a first feature value extracted from an audio signal of a previous frame; analyzing the audio signal of the first frame and extracting a second feature value when the similarity is below a predetermined threshold value; and comparing the extracted first feature value and the second feature value and at least one feature value corresponding to a pre-defined voice signal and determining whether or not the audio signal of the first frame is a voice signal, and thus the electronic device may detect only a voice section from the audio signal while improving the processing speed.

Abstract: A method for speech processing includes determining an unvoicing parameter for a first frame of a speech signal and determining a smoothed unvoicing parameter for the first frame by weighting the unvoicing parameter of the first frame and a smoothed unvoicing parameter of a second frame. The unvoicing parameter reflects a speech characteristic of the first frame. The smoothed unvoicing parameter of the second frame is weighted less heavily when the smoothed unvoicing parameter of the second frame is greater than the unvoicing parameter of the first frame. The method further includes computing a difference, by a processor, between the unvoicing parameter of the first frame and the smoothed unvoicing parameter of the first frame, and determining a classification of the first frame according to the computed difference. The classification includes unvoiced speech or voiced speech. The first frame is processed in accordance with the classification of the first frame.

Abstract: Techniques for creating groups of devices for controlling these groups with voice commands are described herein. For instance, an environment may include an array of secondary devices (or “smart appliances”, or simply “devices”) that are configured to perform an array of operations. Users may request to create different groups of these devices, such that the users may control entire groups at a single time with individual voice commands.

Abstract: Embodiments of the present invention provide an approach for estimating the accuracy of a transcription of a voice recording. Specifically, in a typical embodiment, each word of a transcription of a voice recording is checked against a customer-specific dictionary and/or a common language dictionary. The number of words not found in either dictionary is determined. An accuracy number for the transcription is calculated from the number of said words not found and the total number of words in the transcription.

Abstract: Techniques for defining a set of predefined device capabilities generally offered by available voice-controllable devices are described herein. Thereafter, as a particular user introduces new secondary devices into his environment and registers these devices, the techniques may identify the capabilities of the new device and map these capabilities to one or more of the predefined device capabilities of the set.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for automatic speech recognition using multi-dimensional models. In some implementations, audio data that describes an utterance is received. A transcription for the utterance is determined using an acoustic model that includes a neural network having first memory blocks for time information and second memory blocks for frequency information. The transcription for the utterance is provided as output of an automated speech recognizer.

Abstract: A method and apparatus for speech recognition are provided. The method and the apparatus calculate signal to noise ratios (SNRs) of speech signals from a user received at speech recognition apparatuses. The method and the apparatus recognize a reference speech signal having a maximum SNR among the SNRs.

Abstract: A computer-implemented technique includes receiving, at a computing device including one or more processors, a user input (i) identifying a portion of a media stream being output from the computing device and (ii) indicating a request to translate the portion of the media stream from a source language to a target language. The technique includes transmitting, from the computing device, the portion of the media stream to a translation server via a network in response to receiving the user input. The technique includes receiving, at the computing device, a translated portion of the media stream from the translation server via the network, the translated portion of the media stream having been translated from the source language to the target language by the translation server. The technique also includes outputting, at the computing device, the translated portion of the media stream.

Abstract: Disclosed are a method for determining whether a person is drunk after consuming alcohol capable of analyzing alcohol consumption in a time domain by analyzing a voice, and a recording medium and a terminal for carrying out same.

Abstract: Disclosed are a method for determining whether a person is drunk after consuming alcohol on the basis of a difference among a plurality of formant energy energies, which are generated by applying linear predictive coding according to a plurality of linear prediction orders, and a recording medium and a terminal for carrying out the method.

Abstract: A system of this invention is directed to a speech processing system that efficiently performs noise suppression processing for a plurality of noise sources spreading in a lateral direction with respect to a speaker of interest.

Abstract: An audio encoding apparatus and method that encodes hybrid contents including an object sound, a background sound, and metadata, and an audio decoding apparatus and method that decodes the encoded hybrid contents are provided. The audio encoding apparatus may include a mixing unit to generate an intermediate channel signal by mixing a background sound and an object sound, a matrix information encoding unit to encode matrix information used for the mixing, an audio encoding unit to encode the intermediate channel signal, and a metadata encoding unit to encode metadata including control information of the object sound.

Abstract: Disclosed is a method for determining alcohol consumption capable of analyzing alcohol consumption in a time domain by analyzing a formant slope of a voice signal, and a recording medium and a terminal for carrying out same. An terminal for determining whether a person is drunk comprises: a voice input unit for generating a voice frame by receiving a voice signal; a voiced/unvoiced sound analysis unit for determining whether a received voiced frame corresponds to a voiced sound; a formant frequency extraction unit for extracting a plurality of formant frequencies of the voice frame corresponding to the voiced sound; and an alcohol consumption determining unit for calculating a formant slope between the plurality of formant frequencies, and determining the state of alcohol consumption depending on the formant slope, thereby determining whether a person is drunk by analyzing the formant slope of an inputted voice.

Abstract: Embodiments of the present invention provide an approach for estimating the accuracy of a transcription of a voice recording. Specifically, in a typical embodiment, each word of a transcription of a voice recording is checked against a customer-specific dictionary and/or a common language dictionary. The number of words not found in either dictionary is determined. An accuracy number for the transcription is calculated from the number of said words not found and the total number of words in the transcription.

Abstract: A method of presenting translated content items is disclosed. It is detected that a content item has been captured by a device of a user. It is identified that the content item is a candidate content item for translation. The candidate content item is translated; and the translated candidate content item is presented via a user interface of a wearable display of the device.

Abstract: Systems, methods, and apparatuses are presented for a trained language model to be stored in an efficient manner such that the trained language model may be utilized in virtually any computing device to conduct natural language processing. Unlike other natural language processing engines that may be computationally intensive to the point of being capable of running only on high performance machines, the organization of the natural language models according to the present disclosures allows for natural language processing to be performed even on smaller devices, such as mobile devices.

Abstract: A robust digital fingerprint of a file ensures that one able to produce the robust digital fingerprint has possession of the file. A client obtains information that is unpredictable to the client and uses that information to modify the file and generate a robust digital fingerprint from the modified file. A server, with access to the same unpredictable information, verifies the generated robust digital fingerprint. An algorithm for generating the robust digital fingerprint has a property that different representations of the same content will produce matching digital fingerprints.