Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for training a sequence generation neural network. One of the methods includes obtaining a batch of training examples; for each of the training examples: processing the training network input in the training example using the neural network to generate an output sequence; for each particular output position in the output sequence: identifying a prefix that includes the system outputs at positions before the particular output position in the output sequence, for each possible system output in the vocabulary, determining a highest quality score that can be assigned to any candidate output sequence that includes the prefix followed by the possible system output, and determining an update to the current values of the network parameters that increases a likelihood that the neural network generates a system output at the position that has a high quality score.

Abstract: Aspects of the present disclosure provide systems and methods for access control using multi-factor validation. In an example, an access control system is designed to be used in conjunction with a first recognition system, such as facial recognition system, a gait recognition system, or audio recognition system, that uses a confidence level to determine whether individuals are authorized to access a restricted area. When the first recognition system is unable to confidently identify the individual, a second recognition system, such as a mobile device system or access card system, may be used to provide second factor verification. Further, stored recognition data may be updated to include information gathered by the first recognition system in response to use of the second factor verification.

Abstract: Embodiments described herein include a method for insertion of formatted text with speech recognition. One embodiment of the method includes creating a speech recognition command and executing the speech recognition command. Creating the speech recognition command may include receiving a selection by a user to add a plurality of actions to the speech recognition command and receiving a user-defined trigger command for the speech recognition command. In some embodiments, creating the speech recognition command includes receiving a definition of the text grammar, wherein the definition of the text grammar includes at least one command part and at least one command definition and storing the speech recognition command.

Abstract: The present invention provides a method and system for analyzing human speech during natural language processing interactions between humans and computers to aid in computer learning. The method processes human language tutorial videos each having a visual track, an audio track and captions. Multiple videos are simultaneously processed in parallel using stream processing to identify spoken words or phrases in the videos by comparing them with benchmark words/phrases stored on a computer. Confidence scores are determined for each of the spoken words/phrases which are assigned to a list of the benchmark words/phrases on the computer when a threshold value is met. A system administrator can identify spoken words/phrases to which the threshold value is not met.

Abstract: A method (and system) of determining confusable list items and resolving this confusion in a spoken dialog system includes receiving user input, processing the user input and determining if a list of items needs to be played back to the user, retrieving the list to be played back to the user, identifying acoustic confusions between items on the list, changing the items on the list as necessary to remove the acoustic confusions, and playing unambiguous list items back to the user.

Abstract: Sequential speech recognition using two unequal automatic speech recognition (ASR) systems may be provided. The system may provide two sets of vocabulary data. A determination may be made as to whether entries in one set of vocabulary data are likely to be confused with entries in the other set of vocabulary data. If confusion is likely, a decoy entry from one set of the vocabulary data may be placed in the other set of vocabulary data to ensure more efficient and accurate speech recognition processing may take place.

Abstract: The invention relates to a method and a device for the natural-language recognition of a vocal expression. A vocal expression of a person is detected and converted into a voice signal to be processed by a voice recognition device. Afterwards, the voice signal is analyzed at the same time or sequentially in a plurality of voice recognition branches of the voice recognition device using a plurality of grammars, wherein the recognition process is successfully completed if the analysis of the voice signal in at least one voice recognition branch supplies a positive recognition result.

Abstract: A visualization development system is provided. The system includes a visualization tool to develop one or more visualizations and a grammar engine that operates with the visualization tool to automatically detect visualization problems during the development of the visualizations.