Context-aware news delivery

Delving more deeply into the issue of context-aware news delivery, I wanted to thrash out some ideas about what contexts any given news reader might have, and how we might expose these as configurable variables to help inform my current brainstrain thinking about my MozNewsLab final project.

The main questions are: what can we reasonably detect? How can we use these variables to support the user to filter the news deluge?

device-specific detection

• type - e.g. phone/tablet/laptop/tv
• screen resolution / orientation
• multimedia capabilities e.g. canvas, video, audio (can we detect sound capabilities too?)
• battery life - wouldn't it be handy if your news app noticed you were running low on juice and offered to switch you to a lo-fi version of whatever you're viewing?
• local time/date - if it's early morning where you are, you might want to see summaries of the big stories for the past 8 hours since you've hopefully been tucked up asleep. If it's late at night, you might prefer the latest news only, or deeper context about the day's events. If it's a special occasion (bank holiday / Xmas / whatever is local), you might want special content related to that.
• location - not just giving you the opportunity to see hyperlocal stories, but detecting whether you're on the move, and if so giving you the option to switch to lo-fi mode (text summaries; litte or no graphics; minimal navigation) to preserve mobile bandwidth / help you navigate news on a small screen (can you even do this on webtv?) . Also, what's nearby? Can I pick up more content (e.g. via Bluetooth or NFC) around here? Are there any open calls for contributors of content (video/pics/audio/text) of what's going on near me (similar to Rhiannon's excellent contract webcams concept)?
• proximity - e.g. to NFC devices?

Naturally, we'd want to use something like Modernizr to make detection simpler ... though there's also a bunch of undetectables.

user preferences

• location - if at home, show me x; if on the move, show me y; if at work show me z
• last viewed - might want to pick up where you left off - or have the option to save (and share) snapshots
• frequently viewed - a sort of news macro e.g. at 8am mon-fri, show me travel news from home to work; at 1pm mon-fri, show me entertainment news
• sources - user can add sources; we can suggest sources to them based on their existing choices; possibly share lists of sources?
• content type - also dependant on device - tend to watch audio/video on tv/laptop/tablet; limited nav on tv/mobile means more text, less interactivity?
• friends - what are they reading/watching/sharing/commenting on/producing?

Variables

To respond to these directly measured or inferred indications of context, what can we vary, given a decent amount of metadata for any given news item (official story, image, photo, tweet, status update, google+ post, whatever)?

• subject (would have to be inferred from social media updates, and any non-tagged items, which is potentially a massive overhead)
• location
• format (i.e. text/audio/video/image/interactive thingamajig)
• amount of content (e.g. just headline, full banana)
• author
• source (official/social)

Things we might want to be able to order by

• popularity - most read / most shared / most commented on
• importance - defined officially (by journos) or unofficially (by you or other users)
• date published / updated
• degrees of separation - items from people/organisations you follow to FOAFs to everything else

Stuff we might be able to infer/crowdsource

• political persuasion (needs a corpus of e.g. tweets to try and figure it out)
• sources - can we link a given news item to its related sources? News agencies are starting to quote Twitter more, but they don't tend to link to the tweet (why not? Is it convention, or a CMS template issue?): it would be great if journalists published their sources openly, so every news story has a bunch of rNews hasSource links ... which in turn would enable people to visualise the whole interlinked web of news.

</braindump>