Amazon Technologies, Inc.

Tamojit Chatterjee

Muhammad Raffay Hamid

Mayank Sharma

Sandeep Joshi

Systems, methods, and computer-readable media are disclosed for language-agnostic subtitle drift detection and correction. A method may include determining subtitles and/or captions from media content (e.g., videos), the subtitles and/or captions corresponding to dialog in the media content. The subtitles may be broken up into segments which may be analyzed to determine a likelihood of drift (e.g., a likelihood that the subtitles are out of synchronization with the dialog in the media content) for each segment. For segments with a high likelihood of drift, the subtitles may be incrementally adjusted to determine an adjustment that eliminates and/or reduces the amount of drift, and the drift in the segment may be corrected based on the drift amount detected. A linear regression model and/or human blocks determined by human operators may be used to otherwise optimize drift correction.

Language agnostic drift correction

Systems, methods, and computer-readable media are disclosed for language-agnostic subtitle drift detection and correction. A method may include determining subtitles and/or captions from media content (e.g., videos), the subtitles and/or captions corresponding to dialog in the media content. The subtitles may be broken up into segments which may be analyzed to determine a likelihood of drift (e.g., a likelihood that the subtitles are out of synchronization with the dialog in the media content) for each segment. For segments with a high likelihood of drift, the subtitles may be incrementally adjusted to determine an adjustment that eliminates and/or reduces the amount of drift and the drift in the segment may be corrected based on the drift amount detected. A linear regression model and/or human blocks determined by human operators may be used to otherwise optimize drift correction.

Devices, systems, and methods are provided for language-agnostic subtitle drift detection and localization. A method may include extracting audio from video, dividing the audio into overlapping blocks, and determining the probabilities of overlapping portions of the blocks, the probabilities indicating a presence of voice data represented by the audio in the blocks. The method may generate machine blocks using overlapping portions of blocks where voice data is present, and may map the machine blocks to corresponding blocks indicating that subtitles are available for the video. For mapped blocks, the method may include determining features such as when subtitles are available without voice audio, when voice audio is available without subtitles, and when voice audio and subtitles both are available. Using the features, the method may include determining the probability that the video includes subtitle drift, and the method may include analyzing the video to localize where the subtitle drift occurs.

Growing community of inventors

Shyamnager, India

Tamojit Chatterjee