Mittwoch, 30. September 2009

good relations, twitter & co. dexa 2/2

good relations, a semantic e-commerce ontology was another highlight of dexa. for part one of the coverage, see here. martin hepp, bundeswehr university germany pinpointed to the lack of semantic description of products offered online. this makes search, not only in the long tail, inefficient for sellers and buyers. with good relations he offers a open source basis to tackle this. best.buy and yahoo already use good relations to make their product easier to find. enthusiastically martin hepp was evangelizing for good relations.

in contrast to most of the presenting scientists he focuses on an industry ready product approach. uptake of usage will be interesting to follow.

there have been many other interesting talks, like perspectives of web services intermediaries of urlich scholten et. al or fraud detection by human agents of daniel schwabe.

surprising was the lack of twitter and slideshare. only two other twitterers seemed to be posting. only one presentation can be found online. there is still a way to catch up with non-academic life by those which explore the future.
Reblog this post [with Zemanta]

Montag, 21. September 2009

frogs, elections and e-commerce recommender systems. dexa part 1/2

the 10th edition of the international conference on database expert systems & applications, DEXA 2009, brought four packed days with scientific tracks. though seemingly a tough topic to handle, not everyone can jump on the bandwagon of hardcore tech tracks like „evolution of query optimizations methods...“ for data grid systems of abdelkader hameurlain and the like. so let´s focus more on generic topics and start with the „frogs“.

if one would put a frog in hot water, he´d probably try to jump out. if put in a pot of cold water heated steadily he wouldn´t. although not advised to try that, alessandro acquisti from carnegie mellon university (usa) used this baseline to show recent research on privacy behaviour in the internet.
working in the field of behavioural economics, he tried to find out how the arrangement of questions in surveys would influence the willingness of participants to reveal quite private information. like if people did betray an insurance or tax authorities. following the frog analogy (and thus one common design principle of survey design) people should be more willing to reveal sensitive information about themselves if the questions would steadily get more „private“. acquisti rejected this idea. so here we go, if you want sensitive information from your customers, get to the core right from the beginning.

e-commerce recommender systems have also been proven to be a hugely popular research topic. edith elking from the nanyang technology university in singapore brought up a worthy connection between elections and recommendation system.
recommendation systems have the aim to provide (internet) users with proposals, best fitting to their individual preferences. by an „magic“ accumulation of crowd, social, personal,... data every user should be served with her personal best recommendation. this is a difficult task on many dimensions, like context interpretation or manipulation hardness.

for the latter, most technical scientists try to invent new „bulletproof“ recommendation systems from the scratch. on the other hand have scientists since the french revolution thought about voting systems which should find the best candidate for millions of voters, making manipulations hard, easy to use and to compute.
as arrow´s theorem though states, there is no perfect voting system. based on this aspect of the talk the hypothesis can be stated, that the same will hold for recommendation systems.
it seem´s though that in recommendations systems the technology seems to struggle already with simple majority voting systems.

read more on "good relations", twitter & co in the second part of the dexa coverage.
Reblog this post [with Zemanta]