once i was told that over 160 definition exist for relevance. tefko saracevic has more to tell about it. and he knows the history of that notion in information retrival:

• What are the manifestationsof relevance? • What is the behaviorof relevance? Or more precisely: How do people behave in respect to relevance? • What are the effects of relevance? Or more precisely: What influencesare related to relevance? | Part II: Nature and Manifestations of Relevance*

social networks, cloud computing, multimedia services or search - igual - one has to keep in mind its all about infrastructure. the important questions, according to francois bourdoncle, exalead, for this services are: can i scale? what does the solution cost me? cost of revenue?

bourdoncle further argues that cost of revenues is google's largest asset, together with its brand | ict, 26. november 2008, lyon.

facebook • 13,000 machines in their datacenter • 50,000 machines next year • $1M/month for electricity • $500k/month for network bandwidth • 2TB/day of data (mostly photos) • Add one huge storage system every week • $80M/year for their staff (800) • Burning cash quickly ($516M raised to date) | data according to bourdoncle

twitter is/was known for being down, so no wonder they are in need of money

sitting in a room with a lot of information and interesting people, yet they stay silent. most of the time. left with the feeling i just got to hear the tip of the iceberg. 09wsdm, a split-off of the www conference serial, i was told (this year in madrid), reminded me that there are real world problems of all sort to dig for information. someone told me he spotted greg linden, so i hoped he would write a summary. guess i have to wait.

susan dumais from ms research does interesting work and is quite a source of information.

from her and her colleagues studies you can learn that for more than 40% of clicks don't lead to satisfaction.

ranking: date by far the most common sort field. 50 - 80% of page visits are re-visits | thinking outside the (search)box

google is aiming to translate any language on the web. well, this reminds me of the discussion of arial and unicode: 50.000 glyphs are enough. anyhow, it has been a long time i have heard someone from google excepting questions after his/her talk. a tip of the hat to jeff dean.
interesting historic parts in his talk about the history of google and the development of its data centers: if you have to update that many machines do it at once and not partically; if something goes wrong you harm less people this way.

yahoo is into translating, too. mostly aiming for the ads market.

privatized search has been a topic. take a look at j. teevan, s. dumais and d. liebling: to personalize or not to personalize: modeling queries with variation in user intent.

just realized that i missed the "best student paper" by eytan adar, who worked on zoetrope, teevan, dumais and jonathan l. elsas. well, the paper is online. | via jeff dalton