Abstract: Disclosed herein are systems, methods, and computer-readable storage devices for processing audio signals. An example system configured to practice the method receives audio at a device to be transmitted to a remote speech processing system. The system analyzes one of noise conditions, need for an enhanced speech quality, and network load to yield an analysis. Based on the analysis, the system determines to bypass user-defined options for enhancing audio for speech processing. Then, based on the analysis, the system can modify an audio transmission parameter used to transmit the audio from the device to the remote speech processing system. The audio transmission parameter can be one of an amount of coding, a chosen codec, or a number of audio channels, for example.

Abstract: A system and method are presented for forming the excitation signal for a glottal pulse model based parametric speech synthesis system. The excitation signal may be formed by using a plurality of sub-band templates instead of a single one. The plurality of sub-band templates may be combined to form the excitation signal wherein the proportion in which the templates are added is dynamically based on determined energy coefficients. These coefficients vary from frame to frame and are learned, along with the spectral parameters, during feature training. The coefficients are appended to the feature vector, which comprises spectral parameters and is modeled using HMMs, and the excitation signal is determined.

Abstract: A mechanism is provided updating a knowledge base of a sentiment analysis system, the knowledge base being operable for storing natural language terms and a score value related to each natural language term, the score value characterizing the sentiment of the natural language term. Messages comprising natural language are received. Using content of the knowledge base, a decision is made as to whether at least one message of the received messages has a positive sentiment or a negative sentiment. A term is extracted from the message that is not present in the knowledge base. Based on a frequency of occurrence of the term in the received messages and the sentiment of the messages in which the term occurs, a score value of the term is calculated, and the term and the calculated score value are stored into the knowledge base.

Abstract: Embodiments described herein provide approaches for validating synonyms in ontology driven natural language processing. Specifically, an approach is provided for receiving a user input containing a token, structuring the user input into a semantic model comprising a set of classes each containing a set of related permutations of the token, designating the token as a synonym of one of the set of related permutations, annotating the token with a class from the set of classes corresponding to the one of the set of related permutations, and validating the annotation of the token by determining an accuracy of the designation of the token as a synonym of the one of the set of related permutations. In one embodiment, the accuracy is determined by quantifying a linear distance between the token and a contextual token also within the user input, and comparing the linear distance to a pre-specified linear distance limit.

Abstract: A system and computer implemented method for managing perspective data is disclosed. The method may include collecting a first lot of perspective data for an item. The method may include introducing a variant feature to the item to constitute a modified item. The method may include collecting a second lot of perspective data for the modified item. The method may also include evaluating the first and second lots of perspective data to ascertain a sentiment fluctuation based on information relevant to the variant feature.

Abstract: A computer-implemented method for language generation of a flow diagram, which receives a flow diagram. A plurality of geometric shapes within the flow diagram is identified. A plurality of text elements within the flow diagram is identified. The plurality of text elements and corresponding geometric shapes are associated. The association between the plurality of geometric shapes are identified. A diagram matrix based on the associations between the plurality of geometric shapes is generated. A linear language representation of the diagram matrix is generated.

Abstract: In general, techniques are described for identifying a codebook to be used when compressing spatial components of a sound field. A device comprising one or more processors may be configured to perform the techniques. The one or more processors may be configured to identify a Huffman codebook to use when compressing a spatial component of a plurality of spatial components based on an order of the spatial component relative to remaining ones of the plurality of spatial components, the spatial component generated by performing a vector based synthesis with respect to a plurality of spherical harmonic coefficients.

Abstract: An apparatus and a method for constructing a multilingual acoustic model, and a computer readable recording medium are provided. The method for constructing a multilingual acoustic model includes dividing an input feature into a common language portion and a distinctive language portion, acquiring a tandem feature by training the divided common language portion and distinctive language portion using a neural network to estimate and remove correlation between phonemes, dividing parameters of an initial acoustic model constructed using the tandem feature into common language parameters and distinctive language parameters, adapting the common language parameters using data of a training language, adapting the distinctive language parameters using data of a target language, and constructing an acoustic model for the target language using the adapted common language parameters and the adapted distinctive language parameters.

Abstract: A dialog device comprises a natural language interfacing device (chat interface or a telephonic device), and a natural language output device (the chat interface, a display device, or a speech synthesizer outputting to the telephonic device). A computer stores natural language dialog conducted via the interfacing device and constructs a current utterance word-by-word. Each word is chosen by applying a plurality of language models to a context comprising concatenation of the stored dialog and the current utterance thus far. Each language model outputs a distribution over the words of a vocabulary. A recurrent neural network (RNN) is applied to the distributions to generate a mixture distribution. The next word is chosen using the mixture distribution. The output device outputs the current natural language utterance after it has been constructed by the computer.

Abstract: A collection of data that is extremely large can be difficult to search and/or analyze. Relevance may be dramatically improved by automatically classifying queries and web pages in useful categories, and using these classification scores as relevance features. A thorough approach may require building a large number of classifiers, corresponding to the various types of information, activities, and products. Creation of classifiers and schematizers is provided on large data sets. Exercising the classifiers and schematizers on hundreds of millions of items may expose value that is inherent to the data by adding usable meta-data. Some aspects include active labeling exploration, automatic regularization and cold start, scaling with the number of items and the number of classifiers, active featuring, and segmentation and schematization.

Abstract: A gain control system for controlling gain applied to an audio signal includes a power estimator configured to estimate the power of a digital signal derived from the audio signal, a digital gain estimator configured to determine, in dependence on the estimated power, a digital gain which would modify the power of the digital signal so as to reach a target power level, and a gain controller configured to adjust an analog gain applied to the audio signal in dependence on the determined digital gain.

Abstract: Systems and processes for converting speech-to-text are provided. In one example process, speech input can be received. A sequence of states and arcs of a weighted finite state transducer (WFST) can be traversed. A negating finite state transducer (FST) can be traversed. A virtual FST can be composed using a neural network language model and based on the sequence of states and arcs of the WFST. The one or more virtual states of the virtual FST can be traversed to determine a probability of a candidate word given one or more history candidate words. Text corresponding to the speech input can be determined based on the probability of the candidate word given the one or more history candidate words. An output can be provided based on the text corresponding to the speech input.

Abstract: Aspects of the present disclosure relate to a distributed storytelling framework. A server receives an adjacency list comprising a set of nodes linked together by edges. The server converts the adjacency list to a set of generated storylines, each storyline being represented as a key-value pair. A key represents a first node and a value represents a second node linked to the first node by an edge. The server combines first and second storylines, of the set of generated storylines, to generate an additional storyline in response to a value from a first storyline matching a key from a second storyline. The additional storyline includes a single key and multiple values, and is added to the set of generated storylines. The server repeats combining storylines, of the set of generated storylines, to generate additional storylines. The server provides an output corresponding to at least one of the generated storylines.

Abstract: An electronic device includes a first microphone that receives a sound generated for a specific time period, from the outside, a second microphone, which is disposed at a location spaced apart from the first microphone and which receives the sound, an audio converter comprising audio converting circuitry, and a processor electrically connected with the first microphone, the second microphone, and the audio converter. The processor is configured to convert the sound obtained from the first microphone, into a first signal and to convert the sound obtained from the second microphone, into a second signal, using the audio converter, and to determine the sound, which is generated for the specific time period, as a voice or a noise based on a frequency-related correlation between the first signal and the second signal.

Abstract: A computer-implemented method for language generation of a flow diagram, which receives a flow diagram. A plurality of geometric shapes within the flow diagram is identified. A plurality of text elements within the flow diagram is identified. The plurality of text elements and corresponding geometric shapes are associated. The association between the plurality of geometric shapes are identified. A diagram matrix based on the associations between the plurality of geometric shapes is generated. A linear language representation of the diagram matrix is generated.

Abstract: A system and computer implemented method for managing perspective data is disclosed. The method may include collecting a first lot of perspective data for an item. The method may include introducing a variant feature to the item to constitute a modified item. The method may include collecting a second lot of perspective data for the modified item. The method may also include evaluating the first and second lots of perspective data to ascertain a sentiment fluctuation based on information relevant to the variant feature.

Abstract: According to one embodiment, a machine translation apparatus includes the following elements. The machine translation unit performs machine translation on a first text in a first language to generate a first machine translation result in a second language. The retrieval unit retrieves a first question sentence in the first language similar to the first text to obtain a degree of similarity between the first text and the first question sentence. The determination unit determines a first answer sentence in the first language corresponding to the first question sentence to be an output target when the degree of similarity is higher than a threshold and determines the first machine translation result to be an output target when the degree of similarity is lower than the threshold.

Abstract: In an approach for selectively displaying a push notification, audio is captured using a microphone. A processor receives a push notification, wherein the push notification includes information. A processor identifies a keyword associated with the push notification based on the information. A processor determines that the captured audio includes the keyword. A processor determines whether to display the push notification based on the determination of whether the captured audio includes the keyword.

Abstract: A system and method are presented for forming the excitation signal for a glottal pulse model based parametric speech synthesis system. The excitation signal may be formed by using a plurality of sub-band templates instead of a single one. The plurality of sub-band templates may be combined to form the excitation signal wherein the proportion in which the templates are added is dynamically based on determined energy coefficients. These coefficients vary from frame to frame and are learned, along with the spectral parameters, during feature training. The coefficients are appended to the feature vector, which comprises spectral parameters and is modeled using HMMs, and the excitation signal is determined.

Abstract: Systems, methods, and computer-readable storage devices for performing per-channel automatic speech recognition. An example system configured to practice the method combines a first audio signal of a first speaker in a communication session and a second audio signal from a second speaker in the communication session as a first audio channel and a second audio channel. The system can recognize speech in the first audio channel of the recording using a first model specific to the first speaker, and recognize speech in the second audio channel of the recording using a second model specific to the second speaker, wherein the first model is different from the second model. The system can generate recognized speech as an output from the communication session. The system can identify the models based on identifiers of the speakers, such as a telephone number, an IP address, a customer number, or account number.