Watching EUscreen beta
- Example of how to search, problem of context of a search result. ("Millenium" in program title, vs description, vs. evt. subtitles)
- How to mark Transcrpitions in a standard way <!>
- The question of "related videos": related in what way? How to give a sense of a personal "workspace" via RDF to define / constrain relevance.
- Annotation as an answer to the "content needs to be short" issue
- metadata mapping tool: overview of relations/properties used within a specific collection
- Important to expose/use RDF's language facility in AA
Quote of the day:
Het is een heel mooi tool als je heel veel tijd hebt om alle vakjes in te vullen.
This project is User led / user driven, maar daar zijn we nog niet aan toe.
... but later...
The more I work with microdata, the more I realize that it succeeds at staying close to the markup. My initial idea was that it translated "naturally" to JSON, but in fact this is not true. It's more that it, by design, is kept intimately connected to the DOM, and surprisingly this is a good thing. Whereas with RDF/A, one must frequently work backwards from the library that efficiently strips out the data into pristinely clean triple-graphs, microdata's DOM-oriented API holds you to the document, so that my current "inheritance" code needs to (naturally) follow the hierarchical structure of the DOM. To my surprise, this feels like a very good design choice as it's no longer necessary to "rebuild" a sense of context, it's simply there.
Bootstrapping the wiki
Lessons learned from pmwiki:
Wonderful way of "bootstrapping" settings / parameters based on clever namespaces of wiki pages. (For instance, settings permissions on pages in a group may be done by creating a page in the group with a particular name and setting a "magic variable" ? ).
For AA, how could settings like "autoplay" or preferences for player be set via wiki pages. What "namespaces" need to exist (User / Group / Project) and how to make it efficient for the system to refer to those values.
One possible feature to enable this would be:
(Visual) Markup translates into indexed/indexable values
Sets the "title" value of a section and... turns into
A table establishes a set of name-value pairs connected to the section.
Cached CSS Queries as Data Accessors
The system allows css queries to be used to access pages. These queries can optionally be cached to create dynamic indexes of a collection of resources.
Take the practice of HTML scraping and reflect them back in an RDF/SPARQL world.
Example: Create a table of public clips tagged "Vicarious Learning".
- "TransLearning" style: Clips not tagged "Private", that are tagged "Vicarious Learning". For each list the clip url, start and end time, clip title, and thumbnail.
- "Beyond...": Imagine queries that work against HTML pages / RDF data.
- Give me all images smaller than 250 wide, from all pages on the Tapies Foundation (maybe within a particular class).
This suggests writing a scraper django app! With magic scaper queries!
Lessons from, among other things TransLearning...
- Radical flatness
The separation of "clips" from "transcript" as separate (but related / overlapping) texts as opposed to some false hierarchical document (the transcript is not subordinate to the decision of the clip).
- Text is the interface
- Translate the visual into the textual
- Markup as use, Extrapolate from the use
Stephanie's example of architect making paths where students walked in the snow, to the example of using visual markup (like an h1 label to mark the purpose of an annotation, to establish a "kind" of annotation).
(Like this idea of working backwards from the visual -- ie if annotations could have a color, then later treat a color as a tag). Similarly the idea of translating the position / size of a note to text to eventually later work with as tag.
One thing from TransLearning, the example of "show me a list of names connected to the set of clips I (an "unauthorized" guest" are authorized to view).