Making Metadata Work
In Denver, after continued review of the development spec for the theme recommendation engine, we have some concerns about the amount of meta data to collect from producers. More specifically, we are trying to find the balance in the amount and meta data fields to request and/or require. The stations that have provided feedback have similar concerns.
With a lot of discussion, we have narrowed down the genre list to about nineteen. I have attached another updated genre listing to this post. The listing is also below. I don't present this list with an assumption that it is finalized. I think that we all kind of need to chew it over and figure out what works best to meet our goals. Those goals for the final product are to,
- Simplify the produce experience,
- Create a system where all station control scheduling via theme blocks,
- Facilitate sharing, and
- Create an archive of public access content with accurate metadata.
Using the new list please the relationship between the listed genres and your station's theme blocks.
Listing
Doesn't look like I can attach and xml file to a post of GDO. I will e-mail the list to the listserv. This list is typed too.
Art and Design
Culture/Ethnicity
Education/Schools
Entertainment
Environment & Animals
Film & Animation
Gender/Sexuality
History & Biography
Home/Garden/Food
How-to/Style
Lifestyle/Health
Music
New Age/Spiritual/Religous
News & Politics & Business
Nonprofits & Activism
Science & Technology
Sports/Hobbies
Travel & Events
Youth
We continue to have our
We continue to have our debate here regarding simplicity of collection vs accuracy of metadata and are looking forward to seeing the results of the mapping exercise at other stations.
Building on what John mentioned, our debates are focusing around:
1. At some undefined point, asking for more metadata == less metadata / less accurate metadata received. Ex: Our current ESCORT implementation asks for a great deal of metadata but none of our producers are willing to spend the time filling it out accurately. It's a bell curve, and we're in disagreement internally over where the best balance is. Some of us think it's collecting very little metadata, others are pushing for more.
The more metadata we have, the more accurately we can automatically categorize content received via the sharing system. In Denver, the argument is being made on one end that we have to have at least the reduced topic list + the audience list in order to make the 'recommendation engine' work and automatically place new content into a station's custom theme blocks. On the other side of the argument we have people suggesting that the additional accuracy obtained from the audience list is not great enough to require producers to select from the additional options, and that we should focus on having just the topic list and assume that in any case there will need to be some manual intervention with some shows to make sure they end up in the appropriate theme block at a station. We also have someone here arguing for no topic list, and just mapping theme blocks to theme blocks, making the ingest form for the producer extremely simple.
With initial mapping results, we're starting to see another problematic situation where different stations would map the same show into different items on the topic list. If this happened too frequently, it would make the topic list fairly useless, and would be an argument towards just mapping theme blocks to theme blocks.
Projects vs Shows and metadata entry. With the new metadata system, we are going to make sure that any metadata entered into the project will automatically propogate into shows so producers can avoid double entry. This is great unless a project picks a general topic say 'cultural issues', but has shows that vary significantly on topic like GLBT, environmental, etc. In this case, it is highly likely that a producer would just leave the show metadata the same as the project metadata, at which point having a more fine-grained list of topics/audiences is defeated.
Finally, we are frequently coming into a debate on whether the automated system should play a role as content curator or whether that role should remain largely with the station directors / outreach / human-to-human interaction. If it needs to play the content curation role, then more metadata is necessary -- i.e. if you have a cultural block but want to be making sure that a good chunk of latino/a programming is automatically scheduled into that block, then you would need to have the audience metadata in order to do that. Otherwise a producer that had a show about well being and lifestyle, albeit primarily latino/a focused, is often likely to pick the well being and lifestyle theme block as opposed to the latino/a theme block. The alternative is for a station's staff to interact more with producers, have a greater understanding of the content being ingested, and make manual adjustments to the automated schedule when necessary.
Hopefully I didn't taint any of that with my personal opinions, anyone feel free to correct me if I did!