Improve video topic assignments

[mbertrand]

Description/Context

Video topics are currently based on a MoreLikeThis opensearch query against courses (id, title, description, full_description). The results are often not that great, for example a video on musical instruments, rhythm, and tone was assigned topics "Computer Science", "Mathematics", "Probability and Science".

Plan/Design

Possible options include:

• Try to find a matching course number within the video's title, description, and transcript. If found, assign that course's topics. If not, then use the MoreLikeThis search.
• Add transcript text to the data sent to the opensearch query and see if that leads to better results.