Abstract: A transcription of audio speech included in electronic content associated with a meeting is created by an ASR model trained on speech-to-text data. The transcription is post-processed by modifying text included in the transcription, for example, by modifying punctuation, grammar, or formatting introduced by the ASR model and by changing or omitting one or more words that were included in both the audio speech and the transcription. After the transcription is post-processed, output based on the post-processed transcription is generated in the form of a meeting summary and/or template.

Abstract: A transcription of audio speech included in electronic content associated with a meeting is created by an ASR model trained on speech-to-text data. The transcription is post-processed by modifying text included in the transcription, for example, by modifying punctuation, grammar, or formatting introduced by the ASR model and by changing or omitting one or more words that were included in both the audio speech and the transcription. After the transcription is post-processed, output based on the post-processed transcription is generated in the form of a meeting summary and/or template.

Abstract: A transcription of audio speech included in electronic content associated with a meeting is created by an ASR model trained on speech-to-text data. The transcription is post-processed by modifying text included in the transcription, for example, by modifying punctuation, grammar, or formatting introduced by the ASR model and by changing or omitting one or more words that were included in both the audio speech and the transcription. After the transcription is post-processed, output based on the post-processed transcription is generated in the form of a meeting summary and/or template.

Abstract: According to some embodiments, a multi-layer speech recognition transcript post processing system may include a data-driven, statistical layer associated with a trained automatic speech recognition model that selects an initial transcript. A rule-based layer may receive the initial transcript from the data-driven, statistical layer and execute at least one pre-determined rule to generate a first modified transcript. A machine learning approach layer may receive the first modified transcript from the rule-based layer and perform a neural model inference to create a second modified transcript. A human editor layer may receive the second modified transcript from the machine learning approach layer along with an adjustment from at least one human editor. The adjustment may create, in some embodiments, a final transcript that may be used to fine-tune the data-driven, statistical layer.

Abstract: A transcription of audio speech included in electronic content associated with a meeting is created by an ASR model trained on speech-to-text data. The transcription is post-processed by modifying text included in the transcription, for example, by modifying punctuation, grammar, or formatting introduced by the ASR model and by changing or omitting one or more words that were included in both the audio speech and the transcription. After the transcription is post-processed, output based on the post-processed transcription is generated in the form of a meeting summary and/or template.

Abstract: The present application provides an intelligent cloud traffic control system, which has a traffic data recalling and transmission architecture arranged hierarchically from bottom to top, and which includes: terminal devices, first-level cloud devices, second-level cloud devices, . . . ith-level cloud devices; wherein the ith-level cloud devices include a first cloud device, . . . a jth cloud device; the ith-level cloud devices further include: a first cache cloud, . . . a kth cache cloud arranged corresponding to the jth cloud device; wherein the jth cloud device stores a jth cloud device traffic data set, and the kth cache cloud stores a kth characteristic traffic data set; and data in the kth characteristic traffic data set has a higher priority of being recalled than data in the jth cloud device traffic data set.

Abstract: The present application provides an intelligent cloud traffic control system, which has a traffic data recalling and transmission architecture arranged hierarchically from bottom to top, and which includes: terminal devices, first-level cloud devices, second-level cloud devices, . . . ith-level cloud devices; wherein the ith-level cloud devices include a first cloud device, . . . a jth cloud device; the ith-level cloud devices further include: a first cache cloud, . . . a kth cache cloud arranged corresponding to the jth cloud device; wherein the jth cloud device stores a jth cloud device traffic data set, and the kth cache cloud stores a kth characteristic traffic data set; and data in the kth characteristic traffic data set has a higher priority of being recalled than data in the jth cloud device traffic data set.

Abstract: The present disclosure relates to an intelligent traffic cloud control system; by making full use of technologies such as cloud computing, edge computing, big data, and artificial intelligence, the present disclosure proposes a technical architecture of a cloud-edge-end cooperative road traffic cloud control platform, which can overcome the defects of the technical architecture of existing road traffic operation management and monitoring systems. The highway cloud control platform designed and developed based on this technical architecture can improve the operational efficiency of the road network under the premise of ensuring safety, so that highway users can obtain more timely, effective, and personalized driving and traveling assistance information; moreover, practical applications such as vehicle-road cooperative autonomous driving and other functions are supported.

Abstract: According to some embodiments, a multi-layer speech recognition transcript post processing system may include a data-driven, statistical layer associated with a trained automatic speech recognition model that selects an initial transcript. A rule-based layer may receive the initial transcript from the data-driven, statistical layer and execute at least one pre-determined rule to generate a first modified transcript. A machine learning approach layer may receive the first modified transcript from the rule-based layer and perform a neural model inference to create a second modified transcript. A human editor layer may receive the second modified transcript from the machine learning approach layer along with an adjustment from at least one human editor. The adjustment may create, in some embodiments, a final transcript that may be used to fine-tune the data-driven, statistical layer.

Abstract: Methods, systems, and computer storage media are provided for facilitating data rotation operations using a multi-threaded implementation. In one embodiment, a volume analysis operation is executed to identify a first candidate data set for rotation from a first volume tier of a volume to a second volume tier of the volume associated with a file system. Thereafter, a rotation entry is added in a queue, the rotation entry generally indicates the first candidate data set for rotation from the first volume tier to the second volume tier. A first rotation worker thread can access the queue and rotate the first candidate data set from the first volume tier to the second volume tier, while a second volume analysis operation is executed to identify a second candidate data set to rotate. Further, a second rotation worker thread can access the queue and rotate the second candidate data set.

Abstract: A collimated object is adjustable to produce collimated light propagating along different propagation directions. The plenoptic imaging system under calibration captures plenoptic images of the object adjusted to different propagation directions. The captured plenoptic images includes superpixels, each of which includes subpixels. Each subpixel captures light from a corresponding light field viewing direction. Based on the captured plenoptic images, a calibration module calculates which propagation directions map to which subpixels. The mapping defines the light field viewing directions for the subpixels. This can be used to improve processing of plenoptic images captured by the plenoptic imaging system.

Abstract: A plenoptic camera captures a plenoptic image of an object illuminated by a point source (preferably, collimated illumination). The plenoptic image is a sampling of the four-dimensional light field reflected from the object. The plenoptic image is made up of superpixels, each of which is made up of subpixels. Each superpixel captures light from a certain region of the object (i.e., a range of x,y spatial locations) and the subpixels within a superpixel capture light propagating within a certain range of directions (i.e., a range of u,v spatial directions). Accordingly, optical properties estimation, surface normal reconstruction, depth estimation, and three-dimensional rendering can be provided by processing only a single plenoptic image. In one approach, the plenoptic image is used to estimate the bidirectional reflectance distribution function (BRDF) of the object surface.

Abstract: A collimated object is adjustable to produce collimated light propagating along different propagation directions. The plenoptic imaging system under calibration captures plenoptic images of the object adjusted to different propagation directions. The captured plenoptic images includes superpixels, each of which includes subpixels. Each subpixel captures light from a corresponding light field viewing direction. Based on the captured plenoptic images, a calibration module calculates which propagation directions map to which subpixels. The mapping defines the light field viewing directions for the subpixels. This can be used to improve processing of plenoptic images captured by the plenoptic imaging system.

Abstract: A collimated object is adjustable to produce collimated light propagating along different propagation directions. The plenoptic imaging system under calibration captures plenoptic images of the object adjusted to different propagation directions. The captured plenoptic images includes superpixels, each of which includes subpixels. Each subpixel captures light from a corresponding light field viewing direction. Based on the captured plenoptic images, a calibration module calculates which propagation directions map to which subpixels. The mapping defines the light field viewing directions for the subpixels. This can be used to improve processing of plenoptic images captured by the plenoptic imaging system.

Abstract: A collimated object is adjustable to produce collimated light propagating along different propagation directions. The plenoptic imaging system under calibration captures plenoptic images of the object adjusted to different propagation directions. The captured plenoptic images includes superpixels, each of which includes subpixels. Each subpixel captures light from a corresponding light field viewing direction. Based on the captured plenoptic images, a calibration module calculates which propagation directions map to which subpixels. The mapping defines the light field viewing directions for the subpixels. This can be used to improve processing of plenoptic images captured by the plenoptic imaging system.

Abstract: A plenoptic camera captures a plenoptic image of an object illuminated by a point source (preferably, collimated illumination). The plenoptic image is a sampling of the four-dimensional light field reflected from the object. The plenoptic image is made up of superpixels, each of which is made up of subpixels. Each superpixel captures light from a certain region of the object (i.e., a range of x,y spatial locations) and the subpixels within a superpixel capture light propagating within a certain range of directions (i.e., a range of u,v spatial directions). Accordingly, optical properties estimation, surface normal reconstruction, depth estimation, and three-dimensional rendering can be provided by processing only a single plenoptic image. In one approach, the plenoptic image is used to estimate the bidirectional reflectance distribution function (BRDF) of the object surface.