Victus Spiritus


Best of Day News Filtering

10 Feb 2011

In the quest for relevant knowledge we spend hours each week (or day) sifting through mountains of data in order to discover information critical to our success. Dozens of services specialize and survive by delivering relevant content.* There's a thin line between brilliant success and obsolescent oblivion.

Relevance media businesses navigate precarious paths between peaks of community praise, cliffs of customer frustration, and spam
pits which appear every step of the way thanks to sinister agents who profit from siphoning attention whenever possible. Fortunately for us, the sharpest minds are tackling the organization and filtering of data into practical products which inform decision making, and reduce shouting spammers to nothing more than background noise (props to gmail's spam filter).

Best of Day

Best of day is an idea I first observed implemented in Friendfeed a couple of years ago. It highlighted the updates with the greatest number of likes from people you follow, in the previous day. I propose a valuable query language and relevance product which identifies not only the update with the largest social support in the previous day, but includes weighted voting to give active authorities^ an edge, and further breakdowns results by topic. This would enable dynamic news updates that would answer queries like: (*italicized* words are replaceable)

"what were the *critical* updates about *mobile phones* *today*?"

"what were the *popular* *JavaScript framework* updates discussed *today*?"

"what *startup acquisitions* *greater than* *50 million* *dollars* happened *this month* in *the US*?" 

The signals can be optimized to identify varied human recognizable sentiment such as critical, popular, or emotional moving content (love/hate). This can be done through the identification of social reaction keywords (language) and media presence, and analytics (visits, clicks, time on page). The data requires algorithmic and social processing to attach meaningful meta data structure.


*= Relevance is key to a broad range of web and app services. It's critical to search engines like DuckDuckGo and Google, social/algorithm hybrid filters such as KnowAbout.itMy6Sense, Blekko and Jawaya, news feed readers like Google Reader and Feedly, visual presentation layers like Flipboard and, topical blog curation by Equentia and media hubs like Twitter, Facebook and Yahoo. The list goes on as long as there are services that wish to attract, retain attention, and earn return visits.

^= Identifying authorities is a hard problem that many relevance companies are interested in solving (Klout, Quora, Twitter). The trick is in balancing consistent high quality contributions versus spikes in popularity as well as more complex pain points.