netscan - microsoft
identity, reputation, exchange, incentive
this guy works for microsoft - he is a sociologist
the challenge - offline it is easy to determine: where people are, where they are coming form or going to how they are grouped and roughtly what they are doing.
online is is difficult to determine: the size and composition of the crowd, the distribution of people into clusters and variations in the social character of groups...
so how do you learn about these groups, individuals - you data mine :)
they process and analyse posts - based on use figures.
interesting thing - core group is ~2% of the total population - they calculate this figure based on return ratios.
13.1 million unique authors in 2002 into newsgroups.
microsofts interest is because they want to encourage use of newsgroups for support purposes.
third buffy reference for the day!
netscan.research.microsoft.com
this stuff is stats heaven -
they have an interesting idea that cross postings provide a way of understanding the focus level of the group.
they organise threads according to attributes (ie. they don't care about the content itself) like the aggregation properties - ie. a large thread prolly plus short growth period may mean controversial thread.
they do the same about authors - how long have you participated, how many posts, how many replies, days seen in environment, etc in order to evaluate the value of the author/contributor.
He has just put up the most beautiful picture of a visualisation of a thread...i want it on a t'shirt - it is fantastic. oh another fab image...a visualisation of the whole of newsnet - awesome - looks like a satelite image - cool.
they are trying to infer reputation from some raw numbers...
something is tickling the back of my mind here about microsoft mining this information - I am sensing a dependency model in the making here...they gather (not many others could gather this depth I am guessing?) and we consume and we get hooked on it...in return for our consumption...mmm
some more great visualisations - really cool thinking
they are using the data mined stuff to allow better access to newsgroups.
something weird...object blogging...he is suggesting you scan objects that you an then annotate it
every object has a story -
netscan.research.microsoft.com
Posted by paula at April 25, 2003 12:16 AM