Is RSS dead, and why is it dying?
Wed Jan 14, 2009
From Science of the invisible I picked up the link to the Read Write Web article on death of enterprise RSS. What is going on here? I'm working in the enterprise and I use RSS a hell of a lot. After trying a number of different strategies I now use Google reader, because it has the fastest way for me to get rid of content in the feeds. Why does the enterprise not get it? I think the issue has to be with signal to noise. RSS opens up feeds to a lot of interesting information, but it is hard work to keep up to date with the content coming in through feeds. Reading feeds is rarely directly relevant to tasks at hand within a company. If we think of RSS as a river of information, most people don't swim in it, but rather go fishing every now and again when they need to find a data point or write a report or something similar. There is a lot of data, but data without insight is of little use.
The tools are not currently in place to easily filter and extract intelligence from that information. Best practices currently involve people creating their own personal workflows, and that is probably a step too far for most people. My take is that what is needed is a tool that can do multiple levels of analysis. Google reader, with it's Trends, is starting to get there, but this only shows reading behavior and does little to highlight interesting things in the content that is being read.
What one wants to extract are probably interesting events or facts where in this context interest has to be a user defined filter. Here are examples of the kind of analysis that might be interesting.
- Key word extraction and time evolution of these keywords, either through TFIDF of some other algorithm.
- Key words determined through cues from social tagging.
- Pattern burst detection.
- Feed Feed term correlations and time evolution of these correlations.
- Rare but interesting terms (what is not being talked about).
- Surprising events, where this can be determined through a Bayseian filter looking for changes from a prior distribution within the feed.
- A mechanism for learning based on user preferences.
- A nice visualization of all of this.
All of this should be packaged up with filters for doing this analysis on user read items, all items in user subscribed feeds, feeds similar to user subscribed feeds and feeds subscribed or read by the contextual social network of the user, whether that be class mates, friends, or colleagues on a project.
There are many people working on this, for sure.
The Nature Blogs project and Scintilla already do some of these things. Writing this makes me think I should pick up Programing Collective Intelligence again!