Abstract: One embodiment of the present invention sets forth a technique for generating a summary of a recording. The technique includes generating an index associated with the recording, wherein the index identifies a set of terms included in the recording and, for each term in the set of terms, a corresponding location of the term in the recording. The technique also includes determining categories of predefined terms to be identified in the index and identifying a first subset of the terms in the index that match a first portion of the predefined terms in the categories. The technique further includes outputting a summary of the recording comprising the locations of the first subset of terms in the recording and listings of the first subset of terms under one or more corresponding categories.

Abstract: Embodiments of the present disclosure provide a method and apparatus for processing a multi-language text. According to embodiments of the present disclosure, the multi-language text including contents in a plurality of languages may be encoded with a Unicode. The method further comprises splitting the multi-language text into a plurality of parts based on the Unicode of the multi-language text, contents of the plurality of parts having different languages. In addition, the multi-language text may also be processed based on the plurality of parts.

Abstract: Systems and methods are described for generating a transcript of a legal proceeding or other multi-speaker conversation or performance in real time or near-real time using multi-channel audio capture. Different speakers or participants in a conversation may each be assigned a separate microphone that is placed in proximity to the given speaker, where each audio channel includes audio captured by a different microphone. Filters may be applied to isolate each channel to include speech utterances of a different speaker, and these filtered channels of audio data may then be processed in parallel to generate speech-to-text results that are interleaved to form a generated transcript.

Abstract: Some embodiments provide systems and methods that can alleviate the burdens on professional advisors of preparing snapshots and answering client questions, particularly questions that do not require professional judgment or strategy. Such embodiments can automate creation of snapshots of account data collected from multiple institutions, suggest questions that are relevant to a user's accounts, and provide programmatic responses to factual or definitional questions pertaining to those accounts.

Abstract: Systems and methods for Deep Learning techniques based multi-purpose conversational agents for processing natural language queries. The traditional systems and methods provide for conversational systems for processing natural language queries but do not employ Deep Learning techniques, and thus are unable to process large number of intents. Embodiments of the present disclosure provide for Deep Learning techniques based multi-purpose conversational agents for processing the natural language queries by defining and logically integrating a plurality of components comprising of multi-purpose conversational agents, identifying an appropriate agent to process one or more natural language queries by a High Level Intent Identification technique, predicting a probable user intent, classifying the query, and generate a set of responses by querying or updating one or more knowledge graphs.

Abstract: A method of coding for multi-channel audio signals includes estimating comparison values at an encoder indicative of an amount of temporal mismatch between a reference channel and a corresponding target channel. The method includes smoothing the comparison values to generate short-term and first long-term smoothed comparison values. The method includes calculating a cross-correlation value between the comparison values and the short-term smoothed comparison values. The method also includes adjusting the first long-term smoothed comparison values in response to comparing the cross-correlation value with a threshold. The method further includes estimating a tentative shift value and non-causally shifting the target channel by a non-causal shift value to generate an adjusted target channel. The non-causal shift value is based on the tentative shift value. The method further includes generating, based on reference channel and the adjusted target channel, at least one of a mid-band channel or a side-band channel.

Abstract: A speech processing apparatus includes a specifier, and a modulator. The specifier specifies any one or more of one or more speeches included in speeches to be output, as an emphasis part based on an attribute of the speech. The modulator modulates the emphasis part of at least one of first speech to be output to the first output unit and second speech to be output to the second output unit such that at least one of a pitch and a phase is different between the emphasis part of the first speech and the emphasis part of the second speech.

Abstract: A device control apparatus controls an operation device. A plurality of persons is engaged in operations of the operation device. The device control apparatus comprises a control circuit configured to obtain voice instructions or voice responses from the plurality of persons, the voice instructions or voice responses including modalities of operations of the operation device, classify the modalities included in the voice instructions or voice responses, and determine a preferential operation of the operation device according to consistency between a first modalities classified from the voice instructions or a second modalities classified from the voice responses and priorities predefined for each person of the plurality of persons.

Abstract: A method includes associating, for each one of a plurality of answer definitions, at least one or more pattern-form questions, wherein each answer definition has an associated jump target that defines a respective entry point into the workspace analytics system to provide information responsive to the associated one or more pattern-faun questions. The method further includes receiving a user input including capturing input text defining a natural language user query, matching the received input text to one of the pattern-form questions thereby selecting the jump target associated with the matched pattern-form question, and generating a response to the natural language user query by retrieving information from the workspace analytics system by referencing a link based on the selected jump target and zero or more parameters values.

Abstract: System, method, and computer product embodiments for processing audio commands are described. A multimedia device receives a digitized audio sequence from a control device. The digitized audio sequence is associated with a display command for controlling a display device. Then, the digitized audio sequence is converted into an audio signature that corresponds to the digitized audio sequence. The multimedia device sends the audio signature to a platform server configured to use the audio signature to process the display command. After, the multimedia device receives content information from the platform server resulting from processing the display command. Then, the multimedia device causes the display device to provide media content related to the content information.

Abstract: Described herein are systems and methods for generating natural language sentences with Sequence-to-sequence (Seq2Seq) models with attention. The Seq2Seq models may be implemented in applications, such as machine translation, image captioning, and speech recognition. Performance has further been improved by leveraging unlabeled data, often in the form of a language models. Disclosed herein are “Cold Fusion” architecture embodiments that leverage a pre-trained language model during training. The Seq2Seq models with Cold Fusion embodiments are able to better utilize language information enjoying faster convergence, better generalization, and almost complete transfer to a new domain while using less labeled training data.

Abstract: Temporal regions of a time-based media program that contain spoken dialog in a language that is dubbed from a primary language are identified automatically. A primary language audio track of the media program is compared with an alternate language audio track. Closely similar regions are assumed not to contain dubbed dialog, while the temporal inverse of the similar regions are candidate regions for containing dubbed speech. The candidate regions are provided to a dub validator to facilitate locating each region to be validated without having to play back or search the entire time-based media program. Corresponding regions of the primary and alternate language tracks that are closely similar and that contain voice activity are candidate regions of forced narrative, and the temporal locations of these regions may be used by a validator to facilitate rapid validation of forced narrative in the program.

Abstract: A method includes a voice-activated device establishing a communication channel with a mobile device through a communication interface, receiving a voice command of a user to perform an action, determining, in response to the voice command, the action based at least in part on the voice command, and outputting an audible response corresponding to the determined action. During outputting of the audible response, visual data that includes a representation of the determined action is displayed on the mobile device. The user is enabled to validate or modify the visual data via a user interface of the mobile device.

Abstract: One embodiment of the present invention sets forth a technique for generating a summary of a recording. The technique includes generating an index associated with the recording, wherein the index identifies a set of terms included in the recording and, for each term in the set of terms, a corresponding location of the term in the recording. The technique also includes determining categories of predefined terms to be identified in the index and identifying a first subset of the terms in the index that match a first portion of the predefined terms in the categories. The technique further includes outputting a summary of the recording comprising the locations of the first subset of terms in the recording and listings of the first subset of terms under one or more corresponding categories.

Abstract: A computer evaluates free-form text messages among members of a team, using natural language processing techniques to process the text messages and to assess psychological state of the team members as reflected it the text messages. The computer assembles the psychological state as reflected in the messages to evaluate team collective psychological state. The computer reports a trend of team collective psychological state in natural language text form.

Abstract: Embodiments of the present disclosure can provide systems, methods, and computer-readable medium for providing audio and/or video effects based at least in part on facial features and/or voice feature characteristics of the user. For example, video and/or an audio signal of the user may be recorded by a device. Voice audio features and facial feature characteristics may be extracted from the voice audio signal and the video, respectively. The facial features of the user may be used to modify features of a virtual avatar to emulate the facial feature characteristics of the user. The extracted voice audio features may modified to generate an adjusted audio signal or an audio signal may be composed from the voice audio features. The adjusted/composed audio signal may simulate the voice of the virtual avatar. A preview of the modified video/audio may be provided at the user's device.

Abstract: One embodiment provides a method, including: receiving, at an information handling device, voice input; determining, using at least one sensor associated with the information handling device, whether the voice input comprises voice input provided proximate to the information handling device; and providing, based on determining that the voice input is provided proximate to the information handling device, output responsive to the voice input. Other aspects are described and claimed.

Abstract: Online dictionary extension of word vectors techniques and systems are described that are configured to provide online extension of existing word vector dictionaries and thus overcome the failures of conventional techniques. In one example, a dictionary extension system is employed by a computing system to extend a word vector dictionary to incorporate a new word in an online manner Co-occurrence information is estimated for the new word with respect to the words in the existing dictionary. This is done by estimating co-occurrence information with respect to a large word set based on the existing dictionary and sparse co-occurrence information for the new word. The estimated co-occurrence information is utilized to estimate a new word vector associated with the new word by projecting the estimated co-occurrence information onto the existing word vector dictionary. An extended dictionary is created incorporating the original dictionary and the estimated new word vector.

Abstract: Bandwidth-efficient (i.e., compressed) representations of audio data can be utilized for near real-time presentation of the audio on one or more receiving devices. Persons identified as having speech represented in the audio data can have trained speech models provided to the devices. These trained models can be used to classify the compressed audio in order to improve the quality to correspond more closely to the uncompressed version, without experiencing lag that might otherwise be associated with transmission of the uncompressed audio. The uncompressed audio is also received, with potential lag, and is used to further train the speech models in near real time. The ability to utilize the uncompressed audio as it is received prevents a need to store or further transmit the audio data for offline processing, and enables the further trained model to be used during the communication session.

Abstract: A stereo sound decoding method and system decode left and right channels of a stereo sound signal, using received encoding parameters comprising encoding parameters of a primary channel, encoding parameters of a secondary channel, and a factor ?. The primary channel encoding parameters comprise LP filter coefficients of the primary channel. The primary channel is decoded in response to the primary channel encoding parameters. The secondary channel is decoded using one of a plurality of coding models, wherein at least one of the coding models uses the primary channel LP filter coefficients to decode the secondary channel. The decoded primary and secondary channels are time domain up-mixed using the factor ? to produce the decoded left and right channels of the stereo sound signal, wherein the factor ? determines respective contributions of the primary and secondary channels upon production of the left and right channels.