Abstract

Online video sharing and social media platforms enable viewers to leave comments referring to particular timepoints in a video. However, many times viewers fail to provide a link to the timepoint within the video. This disclosure describes techniques that leverage machine learning models to infer implicit or unmentioned timestamps in a comment on a video and to automatically generate links to the timestamp. Per the techniques, the video is divided into chunks. Each chunk is embedded into a feature space. The comment is optionally broken into segments and each segment embedded is into the same space. Comment segments and video chunks are tested for similarity in embedding space using a similarity measure and matched to each other. The techniques enable video sharing websites and applications (including social media applications) to provide features that enable their users to quickly find sections of videos that match comments of interest to them.

Creative Commons License

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

Share

COinS