Alternative transcription workflow
I'm investigating alternatives to our developed workflow for educators. In contrast to using commercial speech recognition tools, this post examines using freely available tools (namely youtube) for machine transcription, subtitle timecoding, and correction.
1. After uploading your video, go to the "Captions and Subtitles" tab. You can "request processing" to provide machine transcription (using a general speech recognition voice profile on youtube servers) providing you with a basic subtitle "caption track":
2. Editing the subtitles for correction is as simple as pie!
3. The video player provides a synchronised "interactive transcript". Users are able to search for keywords for individual videos, jumping directly to any matching points in the video.
4. Subtitle caption tracks of public videos (see below) are automatically indexed by the search engine. Similar to above, users are able to search across multiple videos and jump directly to matching points.
5. Privacy concerns
There are intellectual-property and commercial-value concerns associated with lecture materials. For sensitive or highly valued course material, educators can restrict the permissions of videos, setting "only those with link can view" (or equivalent) before distributing the restricted URL to enrolled students.
6. Searching restricted videos
Videos with restricted access are sensibly not indexed by the search engine (i.e. users are not able to discover the video URL simply by searching for keywords).
Unfortunately this approach prevents all users (even enrolled students with access to the restricted URLs) from searching across multiple videos for keywords.
A workaround might be to provide access to our transcription indexing tools for trusted users - effectively substituting the youtube search engine with our own.