Abstract: The exemplary embodiments disclose a system and method, a computer program product, and a computer system for determining the intents of user expression. The exemplary embodiments may include receiving a user expression, extracting one or more entities from the user expression, gathering one or more resolvers associated with the one or more entities, identifying a first resolver of the one or more resolvers based on the user expression and the one or more training expressions, and resolving the first resolver to generate a first output.

Abstract: A method for speech level adaptation, the method includes: (A) Receiving, by a voice interactive intelligent personal assistant, multiple input audio signals that includes first and second groups of input audio signals. (B) Buffering the multiple input audio signals. (C) Searching for a voice trigger in the first group of input audio signals. When finding the voice trigger then (D) Determining a linear gain factor to be applied on the second group of buffered audio signals, (E) Applying the linear gain factor on the second group of buffered audio signals to provide the output audio signals; and (F) applying a speech recognition process on the output audio signals to detect an audio command embedded in the output audio signals.

Abstract: Representative embodiments disclose mechanisms to map natural language input to an application programming interface (API) call. The natural language input is first mapped to an API frame, which is a representation of the API call without any API call formatting. The mapping from natural language input to API frame is performed using a trained sequence to sequence neural model. The sequence to sequence neural model is decomposed into small prediction units called modules. Each module is highly specialized at predicting a pre-defined kind of sequence output. The output of the modules can be displayed in an interactive user interface that allows the user to add, remove, and/or modify the output of the individual modules. The user input can be used as further training data. The API frame is mapped to an API call using a deterministic mapping.

Abstract: Disclosed are an apparatus and/or method for providing a sentence based on a user input. When a sentence corresponding to a shorthand word input from a user is generated and provided to the user, a sentence most suitable for a current context is provided to the user by further considering context information. At least a portion of the method for providing a sentence based on a user input may be performed using a rule-based model and/or an artificial intelligence model learned according to at least one of neural network or deep learning algorithms. The rule-based model and/or artificial intelligence model may provide a sentence most suitable for a current context to the user by using the input shorthand word and the context information as input values.

Abstract: Systems and methods for accent and dialect modification are disclosed. Discussed are a method for selecting a target dialect and accent to use to modify voice communications based on a context and a method for selectively modifying one or more words in voice communications in one dialect and accent with one or more vocal features of a different accent.

Abstract: An automatic question answering system serves to respond to questions pertaining to a user-selected domain of a plurality of domains. A plurality of documents pertaining to the domain are received and processed to identify candidate answers. Questions corresponding to the candidate answers are automatically generated. The candidate answers and the questions are used to train a machine comprehension (MC) model. A user interface allows the selection of one of the plurality of domains for querying by a user. When a user query pertaining to the selected domain is received, the user query is analyzed to identify a relevant context from the plurality of documents. An answer span to the user query identified by the MC model from the relevant context is used to frame a complete response to the user query.

Abstract: A scalable system provides automated conversation review that can identify potential miscommunications. The system may provide suggested actions to fix errors in intelligent virtual assistant (IVA) understanding, may prioritize areas of language model repair, and may automate the review of conversations. By the use of an automated system for conversation review, problematic interactions can be surfaced without exposing the entire set of conversation logs to human reviewers, thereby minimizing privacy invasion. A scalable system processes conversations and autonomously marks the interactions where the IVA is misunderstanding the user.

Abstract: An audio system that adjusts one or more beam patterns emitted by one or more loudspeaker arrays based on the preferences of users/listeners is described. The audio system includes an audio receiver that contains a listener location estimator, a listener identifier, and a voice command processor. Inputs from the listener location estimator, the listener identifier, and the voice command processor are fed into an array processor. The array processor drives the one or more loudspeaker arrays to emit beam patterns into the listening area based on inputs from each of these devices. By examining the location, preferred usage settings, and voice commands from listeners, the generated beam patterns are customized to the explicit and implicit preferences of the listeners with minimal direct input. Other embodiments are also described.

Abstract: A natural language processing system includes a model trainer operable to train a natural language processing model based on a training corpus, wherein the training corpus includes a data set from one of multiple languages. An action manager curates application programming interfaces, and a processor receives a natural language query at the natural language processing system, wherein the natural language query is directed to a remote component, and the natural language query is in a language other than English. The processor determines the language of the natural language query, and binds the intent of the natural language query to an action, wherein the action is a function to be carried out by the remote component.

Abstract: There are provided decoding and encoding methods for encoding and decoding of multichannel audio content for playback on a speaker configuration with N channels. The decoding method comprises decoding, in a first decoding module, M input audio signals into M mid signals which are suitable for playback on a speaker configuration with M channels; and for each of the N channels in excess of M channels, receiving an additional input audio signal corresponding to one of the M mid signals and decoding the input audio signal and its corresponding mid signal so as to generate a stereo signal including a first and a second audio signal which are suitable for playback on two of the N channels of the speaker configuration.

Abstract: Embodiments are provided that relate to a computer system, a computer program product, and a machine-implemented method for improving performance of a dialog system employing an automated virtual agent. Embodiments involve clustering traces of dialog conversation logs into sets, and searching the traces for one or more contrastive temporal patterns between the sets.

Abstract: A localized sound projection system can coordinate the sounds of speakers to simulate the placement of an auditory cue in a 3D space. The system can include a plurality of speakers configured to output auditory signals and a sound localization controller configured to control the plurality of speakers to coordinate the auditory signals to simulate an origination location of a patient alarm. The sound localization controller can determine adjusted auditory signals and control a plurality of speakers to output the plurality of adjusted auditory signals. A method for dynamically controlling speaker settings in a medical environment can include determining volume settings corresponding to a speaker, monitoring a level of ambient noise corresponding to a room of a patient, controlling the volume settings of the speaker to reduce or increase a sound level output of a speaker. A patient monitoring system can be configured to physically manipulate medical devices in response to audible commands.

Abstract: A multi-task learning system is provided for speech recognition. The system includes a common encoder network. The system further includes a primary network for minimizing a Connectionist Temporal Classification (CTC) loss for speech recognition. The system also includes a sub network for minimizing a Mean squared error (MSE) loss for feature reconstruction. A first set of output data of the common encoder network is received by both of the primary network and the sub network. A second set of the output data of the common encode network is received only by the primary network from among the primary network and the sub network.

Abstract: A method of processing an audio signal includes receiving an audio bitstream encoded via MPEG Surround 212 (MPS212); generating an internal channel (IC) signal for a single channel pair element (CPE), based on the received audio bitstream, equalization (EQ) values for MPS212 output channels defined in a format converter, and gain values for the MPS212 output channels; and generating stereo output channels, based on the generated IC signal.

Abstract: Systems and methods disclosed herein include, while a microphone of a first networked microphone device is enabled, determining whether a first reference device is in a specific state, and in response to determining that the first reference device is in the specific state, disabling the microphone of the first networked microphone device. Some embodiments further include, while the microphone of the first networked microphone device is enabled, receiving a command to disable the microphone of the first networked microphone device via one of the microphone of the networked microphone device or a network interface of the networked microphone device, and in response to receiving the command to disable the microphone of the networked microphone device via one of the microphone of the networked microphone device or the network interface of the networked microphone device, disabling the microphone of the networked microphone device.

Abstract: Disclosed is an apparatus and method for processing a multichannel audio signal. A multichannel audio signal processing method may include: generating an N-channel audio signal of N channels by down-mixing an M-channel audio signal of M channels; and generating a stereo audio signal by performing binaural rendering of the N-channel audio signal.

Abstract: Embodiments provide a computer implemented method for generating a domain-specific thesaurus on a cognitive system, comprising: receiving data of the domain-specific corpus and a plurality of terms of interest from a user; splitting the data of the domain-specific corpus into a plurality of sentences using natural language processing techniques; for each term in the plurality of terms of interest, retrieving a plurality of candidate sentences containing a corresponding term, from the plurality of sentences; for each candidate sentence, providing a list of synonyms of the corresponding term, wherein the synonyms are contextual alternatives in the corresponding candidate sentence; for each term in the plurality of terms of interest, tracking a frequency of each synonym, and forming a frequency map including all the synonyms of a corresponding term and the frequency of each synonym; and generating a domain-specific thesaurus based on a combination of all the synonyms in the frequency map.

Abstract: This disclosure generally relates to a system and method for assessing reading quality during a reading session. In one embodiment, system is disclosed that analyzes speech that corresponds to a reading session for the duration of the reading session, the consistency of the reading sessions, the speed of the speech during the reading session, the engagement level of the parent during the reading session, and the environment in which the reading session takes place in another embodiment, a method is disclosed for calculating an objective score for a reading session, communicating the score to a parent, and providing suggestions and challenges for improving future reading sessions.

Abstract: Encoding and decoding devices for encoding the channels of an audio system having at least four channels are disclosed. The decoding device has a first stereo decoding component which subjects a first pair of input channels to a first stereo decoding, and a second stereo decoding component which subjects a second pair of input channels to a second stereo decoding. The results of the first and second stereo decoding components are crosswise coupled to a third and a fourth stereo decoding component which each performs stereo decoding on one channel resulting from the first stereo decoding component, and one channel resulting from the second stereo decoding component.

Abstract: Example embodiments disclosed herein relate to filter coefficient updating in time domain filtering. A method of processing an audio signal is disclosed. The method includes obtaining a predetermined number of target gains for a first portion of the audio signal by analyzing the first portion of the audio signal. Each of the target gains is corresponding to a subband of the audio signal. The method also includes determining filter coefficients for time domain filtering the first portion of the audio signal so as to approximate a frequency response given by the target gains. The filter coefficients are determined by iteratively selecting at least one target gain from the target gains and updating the filter coefficient based on the selected at least one target gain. Corresponding system and computer program product for processing an audio signal are also disclosed.