This is going to be discussed in today's project call, but I wanted to get something posted about what we are asking the stations to do over the next two weeks and why since we have so much to cover during this call.
When we started the project, we knew from looking at DOM's data that requiring PBCore genres left us with the majority of videos being classified as Action (the first item on the list) and that open/free tagging didn't work either. While there has been a lot of discussion about moving to a hierarchical genre list since Daniel Westgren initial suggested looking into ESCORT, a few things have become clear.
- Despite great intentions, we don't have the resources to manage a large, organic, hierarchical genre list. I never fully populated the genre list hoping someone would criticize how incomplete some sections were, but that never happened.
- Working with the ACM will not move this forward before we need to have a system for sharing content in place
- Very few producers and/or staff at any of the stations (including DOM) where the existing draft genre list is enabled are using it.
Based on this, John Montgomery has been working with Ann Theis (DOM) and Kate Gorman (UPTV) have been working on restructuring the metadata into multistep process that will result in the recommendation of a Timeslot Theme.
Ann created a spreadsheet comparing Kate's revisions to YouTube, Blip, Academy of Television Arts & Sciences (the Emmys), MNN, and DOM. XML files for the new taxonomies were created that you can import into your site.
The hope would be that over the next two weeks everyone interested could attempt to map these terms to both shows and timeslot themes. We will create some graphical representations of each stations mappings like what we've done here...
http://www.denveropenmedia.org/stats
http://www.denveropenmedia.org/timeslot-themes
After everyone has had a chance to provide feedback, we will update the UI for om_show based on the workflow John has been working on.
Comments
wireframes for the updated UI of om_show
Kevin's been taking a good look our metadata and reported back to us some generalizations. Some of the information we have been looking at can be seen at denveropenmedia.org/stats. Listening to Kevin and reviewing the data, our team came to one conclusion. Producer don't provide accurate metadata and have no incentive to do so.
There are some other repercussions of bad and incomplete metadata:
After talking things through a little bit, we thought that creating an incentive for them to provide accurate and complete metadata would solve the problem. In Denver, we will allow users to choose the first airing of their program as part of the video submission process. That means we will soon be modifying the video submission form to include this scheduling function. So our hypothesis is...
If producers want function of self-schedule and we need accurate and complete metadata, then let's burry the self-scheduling function in the metadata process.
We propose a theme recommendation engine that is essentially the metadata collection process ending at a self-scheduling of the first airing. Stations would create timeslot events, which are tied to timeslot theme (a programmed events on the playback server), and select the metadata and other rules that are tied to the timeslot events. As producer go through the meta data collection process the theme recommendation will try to figure out the best theme for programming. A producer will select the appropriate theme and then be moved to the selef scheduling function.
I've attached a wireframe of the the proposed recommendation engine and included our station's manager thoughts on metadata to be collected through the process. We are not married to what is listed and would like any feedback you may have.
Interesting! Just a quick
Interesting! Just a quick thought, won't producers find it too much to fill in both topic, style, language, target audience and TV rating? I realize the importance of differentiating between topic, style and target audience. But does that make it easier to select a theme, than for the producer to choose the theme themselves?
I guess it depends on how a station work. We have problems having producers to fill in the metadata at all. They rather just want to give us a program and air it when we wish... and they see too much work from their end as an obstacle.
The issue with letting
The issue with letting producers fill out the theme directly is that in many cases, that is all they fill out since it is the only data that influences when the show airs. A few producers completed the other metadata well, but the majority do not. This isn't a problem as long as the content only exists at the station with the Timeslot Theme already selected. It's only when a file is shared with another location that doesn't have a theme that's specific to Denver like Mile High Perspective. Then we need a shared vocabulary to map incoming content to a theme on the receiving station end.
Remember, one of the primary goals on the Open Media Project is to save staff time. If we end up with queues of MPEG2 that requires metadata be manually re-added, we've failed on some level.
In addition to reworking the metadata we are asking for, we are also planning to invest time reworking the UI. Another thing we learned during the beta is that a multiselect list that requires holding the ctrl key does is not intuitive enough for most users.
Yeah, I get your point that
Yeah, I get your point that sharing requires metadata to organize and find the videos by. In Sweden we're not that far yet, but this should be the way we should aim at.