Abstract: Among other things, one or more techniques and/or systems are provided for building an action catalogue, generating an action frame for an action within the action catalogue, and/or executing an action. In an example, an action may be included within the action catalogue based upon descriptive text associated with an application indicating that the application is capable of performing the action (e.g., a movie app may be capable of performing an order movie tickets action). A parameter (e.g., a movie name) and/or an execution endpoint (e.g., a uniform resource identifier used to access movie ticket ordering functionality) may be used to generate an action frame for the action. In this way, user intent to perform an action may be identified from user input (e.g., a spoken command), and the action may be performed (e.g., on behalf of the user with minimal additional user input) by using the action frame.

Abstract: A method and system for summarizing messages from a message stream is disclosed in which association analysis is applied to stream of short data messages comprising words in a spoken language, such as English. Clusters of words are identified that provide a summary of the several conversations (short data messages originating from different human sources) that are imbedded in the message stream. Each word cluster may represent a set of messages that are its instances. The word clusters may collectively constitute a summary of the entire message stream. The word clusters that have been extracted from message stream may also be grouped into topics. Also, an identity of one or more message originators may be listed based on their influence on the messages being analyzed. The short data messages may also be sorted based on a geographical location of one or more originators of messages.

Abstract: A method including: receiving, on a computer system, a text search query, the query including one or more query words; generating, on the computer system, for each query word in the query, one or more anchor segments within a plurality of speech recognition processed audio files, the one or more anchor segments identifying possible locations containing the query word; post-processing, on the computer system, the one or more anchor segments, the post-processing including: expanding the one or more anchor segments; sorting the one or more anchor segments; and merging overlapping ones of the one or more anchor segments; and searching, on the computer system, the post-processed one or more anchor segments for instances of at least one of the one or more query words using a constrained grammar.

Abstract: A computer-implemented method can include receiving a speech input representing a question, converting the speech input to a string of characters, and obtaining tokens each representing a potential word. The method can include determining one or more part-of-speech (POS) tags for each token and determining sequences of the POS tags for the tokens, each sequence of the POS tags including one POS tag per token. The method can include determining one or more parses for each sequence of the POS tags for the tokens and determining a most-likely parse and its corresponding sequence of the POS tags for the tokens to obtain a selected parse and a selected sequence of the POS tags for the tokens. The method can also include determining a most-likely answer to the question using the selected parse and the selected sequence of the POS tags for the tokens and outputting the most-likely answer.

Abstract: An audio coding device that performs predictive coding on a third-channel signal included in a plurality of channels in an audio signal according to a first-channel signal and a second-channel signal, which are included in the plurality of channels, and to a plurality of channel prediction coefficients included in a coding book, the device includes a processor; and a memory which stores a plurality of instructions, which when executed by the processor, cause the processor to execute, selecting channel prediction coefficients corresponding to the first-channel signal and the second-channel signal so that an error, which is determined by a difference between the third-channel signal before predictive coding and the third-channel signal after predictive coding, is minimized; and controlling the first-channel signal or the second-channel signal so that the error is further reduced.

Abstract: An approach is provided to receive audible speech and convert the received speech to text while the audible speech is being delivered to a user. An annotation candidate is identified in the text and an annotation reference relating to the identified annotation candidate is retrieved and presented to the user.

Abstract: A method of deriving speech synthesis parameters from an input speech audio signal, wherein the audio signal is segmented on the basis of estimated positions of glottal closure incidents and the resulting segments are processed to obtain the complex cepstrum used to derive a synthesis filter. A reconstructed speech signal is produced by passing a pulsed excitation signal derived from the position of the glottal closure incidents through the synthesis filter, and compared with the input speech audio signal. The pulse excitation signal and the complex cepstrum are then iteratively modified to minimize the difference between the reconstructed speech signal and the input speech audio signal, by optimizing the position of the pulses in the excitation signal to reduce the mean squared error between the reconstructed speech signal and the input speech audio signal, and recalculating the complex using the optimized pulse positions.

Abstract: Methods and systems for online incremental adaptation of neural networks using Gaussian mixture models in speech recognition are described. In an example, a computing device may be configured to receive an audio signal and a subsequent audio signal, both signals having speech content. The computing device may be configured to apply a speaker-specific feature transform to the audio signal to obtain a transformed audio signal. The speaker-specific feature transform may be configured to include speaker-specific speech characteristics of a speaker-profile relating to the speech content. Further, the computing device may be configured to process the transformed audio signal using a neural network trained to estimate a respective speech content of the audio signal. Based on outputs of the neural network, the computing device may be configured to modify the speaker-specific feature transform, and apply the modified speaker-specific feature transform to a subsequent audio signal.

Abstract: Provided are techniques for providing annotations for revising a message. A message to be sent from a sender to a recipient is received. A meaning map associated with the sender and a meaning map associated with the recipient are obtained. The message is parsed into sub-constructs. The sub-constructs are compared in the meaning map associated with the sender and the meaning map associated with the recipient. Alternative language for the sub-constructs is identified. Annotations are provided based on the alternative language.

Abstract: Methods, systems and computer programs for automatic, highly accurate machine comprehension of a plurality of segments of free form unstructured text in a natural language. The system answers a plurality of complex, free-form questions asked in a natural language, based on the totality of input text. The system further uses a multi-dimensional data model to measure the total effects of actions/verbs acting on various unique nouns present in the input text. The system may convert the questions into another multi-dimensional data model and may then compare the two data models in program memory to derive the answers to the posed questions. The system may then automatically detect unknown words and optionally look them up in digital information sources, such as online dictionaries and encyclopedias, to fill in the gaps in knowledge to answer the questions with expert-like reliability.

Abstract: A system and method of verifying the identity of an authorized user in an authorized user group for enabling secure access to one or more services via a device includes receiving first voice information from a speaker through the device, calculating a confidence score based on a comparison of the first voice information with a stored voice model associated with the authorized user and specific to the authorized user, interpreting the first voice information as a specific service request, identifying a minimum confidence score for initiating the specific service request, determining whether or not the confidence score exceeds the minimum confidence score, and initiating the specific service request if the confidence score exceeds the minimum confidence score.

Abstract: A voice signal processor detects background noise sections to reflect characteristics of the background noise on the Wiener filter coefficient to be used for suppressing noise components of input voice signals. In the voice signal processor, directivity signal generators form directivity signals having a directivity pattern. The directivity signals are used by a coherence calculator to obtain coherence, which is in turn used by a targeted voice section detector to detect a targeted voice section. A background noise section detector detects background noise sections containing no voice signal. When a background noise section is detected, a WF adapter uses characteristics of background noise in the detected temporal section to calculate a new WF coefficient.

Abstract: An encoding method executed by a computer, the method includes converting by the computer information about a transient included in a low-frequency component of an audio signal into information about a transient included in a high-frequency component of the audio signal, detecting, by the computer the transient of the high-frequency component of the audio signal based on the high-frequency component of the audio signal and on the information about the transient of the high-frequency component obtained by the converting; and encoding, by the computer the high-frequency component of the audio signal based on the transient detected by the detecting.

Abstract: Methods, systems, and apparatus, for pattern recognition. One aspect includes a pattern recognizing engine that includes multiple pattern recognizer processors that form a hierarchy of pattern recognizer processors. The pattern recognizer processors include a child pattern recognizer processor at a lower level in the hierarch and a parent pattern recognizer processor at a higher level of the hierarchy, where the child pattern recognizer processor is configured to provide a first complex recognition output signal to a pattern recognizer processor at a higher level than the child pattern recognizer processor, and the parent pattern recognizer processor is configured to receive as an input a second complex recognition output signal from a pattern recognizer processor at a lower level than the parent pattern recognizer processor.

Abstract: In one embodiment, collecting a plurality of words from texts submitted by one or more users; for each of a plurality of communication categories, determining a usage frequency of each of one or more of the words within the communication category based on the texts; and constructing one or more customized dictionaries that each comprise a different blending of selected words.

Abstract: In one embodiment, constructing one or more customized dictionaries for a particular user, each of the customized dictionaries comprising a different blending of one or more frequently used words collected from texts submitted by one or more users; and in response to the user inputting text to an electronic device, selecting one of the customized dictionaries and utilizing it to aid the particular user in inputting text.

Abstract: A system, method, and computer program are provided for identifying message content to send to users based on the users' language characteristics. Language characteristics are extracted from user-generated content and language characteristic scores are assigned to each user. The users are clustered into groups using the language characteristic scores. The system sends test messages with different message content to at least a subset of each group's users and the response rates are measured. For each group, a message content to which the group is most responsive is identified and is associated with the group. Language characteristics from a new user's user-generated content are extracted and language characteristic scores are assigned to the new user. The group to which the new user belongs is identified using the new user's language characteristic scores. A message is sent to the new user with the message content previously associated with the identified group.

Abstract: According to one embodiment, approximate named-entity extraction from a dictionary that includes entries is provided, where each of the entries includes one or more words. Words are read from the entries of the dictionary, and network resources are searched to determine a frequency of occurrence of the words on the network resources. In view of the frequency of occurrence of the words located on the network resources, domain relevancy of the words in the entries of the dictionary is determined. A domain repository is created using top-ranked words as determined by the domain relevancy of the words. In view of the domain repository, signatures for both the entries of the dictionary and strings of an input document are computed. The strings of the input document are filtered by comparing the signatures of the strings against the signatures of the entries to identify approximate-match entity names.

Abstract: A system and method of verifying the identity of an authorized user in an authorized user group through a voice user interface for enabling secure access to one or more services via a mobile device includes receiving first voice information from a speaker through the voice user interface of the mobile device, calculating a confidence score based on a comparison of the first voice information with a stored voice model associated with the authorized user and specific to the authorized user, interpreting the first voice information as a specific service request, identifying a minimum confidence score for initiating the specific service request, determining whether or not the confidence score exceeds the minimum confidence score, and initiating the specific service request if the confidence score exceeds the minimum confidence score.

Abstract: Features are disclosed for managing the use of speech recognition models and data in automated speech recognition systems. Models and data may be retrieved asynchronously and used as they are received or after an utterance is initially processed with more general or different models. Once received, the models and statistics can be cached. Statistics needed to update models and data may also be retrieved asynchronously so that it may be used to update the models and data as it becomes available. The updated models and data may be immediately used to re-process an utterance, or saved for use in processing subsequently received utterances. User interactions with the automated speech recognition system may be tracked in order to predict when a user is likely to utilize the system. Models and data may be pre-cached based on such predictions.