Abstract: A method includes receiving training data that includes a set of transcribed speech utterances where each respective transcribed speech utterance is paired with a corresponding transcription. For each respective transcribed speech utterance, the method includes generating an encoded audio representation and an encoded textual representation, generating a higher order audio feature representation for a corresponding encoded audio representation, generating a higher order textual feature representation for a corresponding encoded textual representation, and determining a loss for the respective transcribed speech utterance based on the higher order audio feature representation and the higher order textual feature representation. The method also includes training a speech encoder and a text encoder of a correction model based on the loss determined for each transcribed speech utterance of the set of transcribed speech utterances.

Abstract: The present application relates to a Bluetooth communication method and a device. The method may be performed by a first device. The first device may establish Bluetooth connection with a second device. The method may include: performing link communication with the second device according to a first modulation mode; in a case where it is detected that the quality of link communication is less than a preset threshold, adjusting the modulation mode of the first device from the first modulation mode to a second modulation mode, and performing the link communication with the second device according to the second modulation mode. The anti-interference capability of the second modulation mode is greater than that of the first modulation mode.

Abstract: A method includes receiving training data that includes a set of transcribed speech utterances where each respective transcribed speech utterance is paired with a corresponding transcription. For each respective transcribed speech utterance, the method includes generating an encoded audio representation and an encoded textual representation, generating a higher order audio feature representation for a corresponding encoded audio representation, generating a higher order textual feature representation for a corresponding encoded textual representation, and determining a loss for the respective transcribed speech utterance based on the higher order audio feature representation and the higher order textual feature representation. The method also includes training a speech encoder and a text encoder of a correction model based on the loss determined for each transcribed speech utterance of the set of transcribed speech utterances.

Abstract: A method includes receiving training data that includes a set of transcribed speech utterances where each respective transcribed speech utterance is paired with a corresponding transcription. For each respective transcribed speech utterance, the method includes generating an encoded audio representation and an encoded textual representation, generating a higher order audio feature representation for a corresponding encoded audio representation, generating a higher order textual feature representation for a corresponding encoded textual representation, and determining a loss for the respective transcribed speech utterance based on the higher order audio feature representation and the higher order textual feature representation. The method also includes training a speech encoder and a text encoder of a correction model based on the loss determined for each transcribed speech utterance of the set of transcribed speech utterances.

Abstract: A split shaker for an incubator, the incubator defining an incubation chamber, wherein the split shaker includes a disk drive motor for driving a shaking table, a disk drive motor including a stator assembly disposed outside the incubation chamber, and a rotor assembly disposed inside the incubation chamber having a central shaft and a rotor turntable mounted on the central shaft; wherein the rotation of the rotor turntable drives a shaking movement of the shaking table.

Abstract: In one embodiment, a method receives a video and information for entities that appear in the video. The video is played in a media player without displaying a queue configured to display one or more entities. When an input is received to display the queue while playing the video, the method performs: determining a set of entities in relation to a time associated with playing the video using the information for the entities that appear in the video and displaying the set of entities in the queue, wherein the video continues to play in the media player while the queue is displayed.

Abstract: In one embodiment, a method generates a plurality of sub-relevance tables including a first set of relevance values between media programs. Each table models relevance values for a single feature in a plurality of features. Labeling results are received that include a second set of relevance values between the media programs. The method combines the sub-relevance tables into a single relevance table that includes a third set of relevance values between the media programs for the plurality of features. The combining generates weights for each of the sub-relevance tables based on the second set of relevance values for the labeling results and the first set of relevance values of the sub-relevance tables that are used to generate the third set of relevance values. A recommendation is provided to a user using the third set of relevance values from the single relevance table and a characteristic of the user.

Abstract: In one embodiment, a method determines a video including ad slots inserted within the video. The method generates a caption curve for a caption file of caption segments for a video based on start and stop times for caption segments in the caption file. The caption segments in the caption file were generated for the video without including ad slots. Then, the method determines a speech velocity for the video using the caption file and revises the caption curve based on the speech velocity and a number of characters in caption segments in the caption file. A speech probability curve is determined based on audio of the video and the method correlates the speech probability curve to the revised caption curve to align the caption segments of the caption file with speech of the video.

Abstract: In one embodiment, a method receives a video and information for entities that appear in the video. The video is played in a media player without displaying a queue configured to display one or more entities. When an input is received to display the queue while playing the video, the method performs: determining a set of entities in relation to a time associated with playing the video using the information for the entities that appear in the video and displaying the set of entities in the queue, wherein the video continues to play in the media player while the queue is displayed.

Abstract: In one embodiment, a method receives a video for a media program and a set of captions for a dialog in the video. A media player plays the video. A time associated with playing of the video is determined and then the method determines a set of entities in relation to the time. The set of entities are included in one or more captions in the set of captions. The method displays the set of entities in a queue where the set of entities are associated with additional information for each respective entity in the set of entities.

Abstract: Particular embodiments provide a watch list of shows to users. The watch list is personalized for each user. Also, the watch list is dynamically organized to predict an order the user will want to watch the shows. Particular embodiments analyze historical user behavior with respect to the timing for recurring releases of the episodes for shows to determine the order of the shows in the watch list. The watch list is organized in a way that a user may select a “watch all” button where unseen episodes for the shows in the watch list are all played to the user in an order that is predicted to be the order in which the user would want to watch the shows. Providing the watch all button makes it important to predict the order of the shows accurately.

Abstract: In one embodiment, a method generates a plurality of sub-relevance tables including a first set of relevance values between media programs. Each table models relevance values for a single feature in a plurality of features. Labeling results are received that include a second set of relevance values between the media programs. The method combines the sub-relevance tables into a single relevance table that includes a third set of relevance values between the media programs for the plurality of features. The combining generates weights for each of the sub-relevance tables based on the second set of relevance values for the labeling results and the first set of relevance values of the sub-relevance tables that are used to generate the third set of relevance values. A recommendation is provided to a user using the third set of relevance values from the single relevance table and a characteristic of the user.

Abstract: In one embodiment, a method determines a video including ad slots inserted within the video. The method generates a caption curve for a caption file of caption segments for a video based on start and stop times for caption segments in the caption file. The caption segments in the caption file were generated for the video without including ad slots. Then, the method determines a speech velocity for the video using the caption file and revises the caption curve based on the speech velocity and a number of characters in caption segments in the caption file. A speech probability curve is determined based on audio of the video and the method correlates the speech probability curve to the revised caption curve to align the caption segments of the caption file with speech of the video.

Abstract: Particular embodiments provide a watch list of shows to users. The watch list is personalized for each user. Also, the watch list is dynamically organized to predict an order the user will want to watch the shows. Particular embodiments analyze historical user behavior with respect to the timing for recurring releases of the episodes for shows to determine the order of the shows in the watch list. The watch list is organized in a way that a user may select a “watch all” button where unseen episodes for the shows in the watch list are all played to the user in an order that is predicted to be the order in which the user would want to watch the shows. Providing the watch all button makes it important to predict the order of the shows accurately.

Abstract: In one embodiment, a method determines a first local binary pattern for a first image in a video and a second local binary pattern for a second image in the video. Then, the method determines an optical flow between the first image and the second image based on a distance between the first local binary pattern and the second local binary pattern. The optical flow is output for use in aligning the first image to the second image.

Abstract: A method for annotating general objects contained in video content is provided. The method sends video data to a client device and receives a first annotation from the client device defining a boundary around a portion of a first frame of the video data. Then, the first annotation is tracked through multiple frames of the video content. Other annotations determined to be associated with annotation that match the first annotation within a threshold are determined where the other annotations are received from other client devices and located in the first frame or other frames from the first frame. The method combines the other annotations and the first annotation into an object track and associates a tag with the object track. The tag is input by at least one of the client devices.

Abstract: In one embodiment, a method receives a video for a media program and a set of captions for a dialog in the video. A media player plays the video. A time associated with playing of the video is determined and then the method determines a set of entities in relation to the time. The set of entities are included in one or more captions in the set of captions. The method displays the set of entities in a queue where the set of entities are associated with additional information for each respective entity in the set of entities.

Abstract: In one embodiment, a method determines a first local binary pattern for a first image in a video and a second local binary pattern for a second image in the video. Then, the method determines an optical flow between the first image and the second image based on a distance between the first local binary pattern and the second local binary pattern. The optical flow is output for use in aligning the first image to the second image.

Abstract: In one embodiment, a method determines multiple screens of multiple mobile computing devices should be combined in playback of a video. A first mobile computing device receives the video and determines device characteristics based on a positioning of the first mobile computing device in relation to a second mobile computing device. Playback characteristics are determined based on the device characteristics. Then, the first mobile computing device renders a first portion of the video on a first screen based on the playback characteristics where a second portion of the video is rendered on a second screen of the second mobile computing device.

Abstract: In a method of processing a super-resolution target image from a plurality of substantially low resolution auxiliary frames, the target image is partitioned into a plurality of adaptively sized blocks, which are sized based upon registration confidence levels of the blocks obtained from information contained in the plurality of auxiliary frames. The blocks are classified into a plurality of different categories according to one or both of their respective registration confidence levels and their respective variance levels. In addition, separate enhancement modes designed to enhance the blocks are selected according to their respective classifications and applied on the blocks to enhance the target image.