The application of computer vision and techniques that rely on machine intelligence has improved automated determination of notable segments within video content. However, further refinement in these techniques can improve the quality and accuracy of the marked segments. Also, it can be difficult for automated techniques to surface the most notable text comments associated with video content. This disclosure describes techniques that combines initial independent automated determination of notable video segments and text comments to refine and improve each aspect based on the other. The techniques leverage audience engagement with live and pre-recorded video content as captured via user comments entered while the video is being viewed. Machine learning models that identify salient video segments can be trained using the output of models that identify notable comments and vice versa, using reinforcement learning.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.