Trip Report WWW2005 2nd day (thursday, may 12)

keynote by eric brewer:

eric brewer's keynote was titled "the case for technology for developing regions". according to moore's law, the processing speed of computers doubles every 18 months, because the number of transistors on a chip doubles every 18 months. this does say nothing about the price of ITC components, but eric believes, that we are now at a point, where ITC is no longer too expensive for developing regions, but we must help them to adopt and use new technologies. it is also important to have the right business models. he gave a number of examples, including the "phone ladies": they get a loan in order to buy a mobile phone for lets say around USD 200. then they let people in their village use their phone for a small fee. this allows them to make a living and to pay back the loan over time. there are other examples of what he referred to as "micro-loans". eric wants to convince high-tech researchers that technology for developing regions is an important and viable research topic.

for more information, see tier.cs.berkeley.edu

(click on the images for an enlargement)

to the top

the Infocious Web search engine:

the goal of the Infocious search engine is to resolve ambiguities of the natural languages by applying extensive language analysis to the Web pages they index. the major components include:

1. the crawler: follows the links to discover Web pages.
2. linguistic processing: resolves language ambiguities, annotates Web pages, performs word analysis (assigns word types such as noun, verb etc), categorization (example "jaguar": could be a car, an animal, a MAC OS, a chemical software), concepts = phrase identification (word combinations identified through a set of rules (pre-specified or automatically learned), e.g. salad dressing) and applies probabilities.
3. text quality: promotes well-written pages

Infocious also provides some enhanced features, such as key phrases, related topics, suggestions for more specific querries and spell checking.

see proceedings for details ...
see also www.infocious.com

to the top

keynote by lorrie faith cranor:

the afternoon keynote was given by lorrie faith cranor from carnegie mellon university and was titled "towards usable Web privacy and security". she claimed that privacy and security nowadays is way too complicated for the average user. it takes too much time and troubles to protect a computer with a firewall, anti-virus software, anti-spyware software, cookie manager, pop-up blocker, spam filter, encryption software etc (see slide below on the right).

approaches to usable security should include software that works, making security and privacy understandable plus some user training. usually developers know more about security than users do, so they should make decisions for them instead of providing e.g. installation options that users don't understand and therefore cannot make the right choices. or as the manager of SP2 for Windows XP stated: present choices, not dilemmas !

passwords are a problem on its own. users must have so many different passwords, that they cannot memorize them all. one possible solution is a password protected password database, so the user must remember only one "super password" in order to access all the other passwords. another option could be a "single sign-on", but this usually works only within one company or organization. another interesting approach are graphical passwords, where a user must either draw a particular image, or she or he must either remember a number of positions on an image or a particular series of pictures. lorrie proposed that at least sites where we log in only occasionally use the already widely available option of sending the password to a pre-registered email address.

lorrie provided a number of examples which showed that security metaphors and symbols are often not very intuitive.

furthermore, she suggested that the use of cookies should be reconsidered: users do not understand the current implementations and implications of cookies. user agents should provide a more sophisticated cookie manager. lorrie suggested to assign cookies to bookmarks and that a cookie manager should support multiple instances, so that a user can have multiple identities for the same website.

many websites have privacy policies posted, but only few were read. they are hard to read and understand. even though P3P is a W3C recommendation since 2002, still the current browser provide only limited P3P functionality. lorrie suggested to make privacy visible on search engines, like her privacy bird does, see privacybird.com for more information.

(click on the images for an enlargement)

to the top

how search engines shape the Web:

this panel discussion was about how search engines affected and still affect the Web. the mechanism of link popularity which boots already well known websites was questioned in that it kind of excludes new or unknown websites from being found. but it was argued, that to provide the most popular document for an ambiguous query is probably the best policy. in addition, if a website exists, one can find it, but it may require more effort.
one panelist explained how some (bad) people try to get more attention from search engines with so called "search engine optimization", also known as "search engine spaming" and how search engine providers try to reduce the impact of such "optimizations". one attendee argued, if his competitors spam search engines, he has to do some optimizations too if he wants his website to be found. because if users don't find his website, he will be soon out of business.
one suggested enhancement was the so called "personalized search", which stores previously found results in your personal history - but this may introduce some privacy issues.
one panelist said, without search engines, the Web would not be possible (or at least would not be as popular as it is today), because there is no incentive in creating content unless it can be found the Web. to use the Web in general and search engines in particular has become the "new normal" in many people's life.
it was stated, that all popular search engines are excellent in navigational queries, which creates false expectations for informational queries, which do not work well at all. also search engines tend to have "false memory" - they remember pages that are no longer accurate.

to the top

disambiguating Web appearances of people in a social network:

this session was about how to find a particular person on the Web if many unrelated persons with the same name have a Web presence. two models were presented. one is the so called "link structure model", which is based on the fact that web pages of a group of people are often linked to each other while the pages of two persons with the same name but no relation are very unlikely interconnected.
the "agglomerative/conglomerative double clustering (A/CDC) model" was also outlined and explained based on some real world examples.
the good news from my point of view is the fact, that the name "reto ambühler" is unique on this planet - whenever you come across a webpage that mentions this name, it's me ... :-)

see proceedings for details ...

to the top

conference dinner:

the conference dinner took place at the hotel new otani just next to the convention center. they provided a large variety of food, but unfortunately no tables to site down. there was also an interesting performance by some japanese musicians and a furious warrior.

(click on the images for an enlargement)

to the top

to the trip report main page