window office

Jun
8th

Mon

Confmaster sucks

Every conference I’ve submitted papers to recently that’s used Confmaster has completely buckled under the pressure of submission day. I’ve been trying to upload the final version of a CIKM submission for over an hour, and all I’m getting is a stuck progress meter. This is really unacceptable.

I maintained the Confmaster system for SIGIR 05, and the system sucks from the administration end, too.

Comments (View)

May
26th

Tue

Just received my first request for a paid link on my website (I assume its my academic homepage). The funny thing is, the requested link is actually for a search engine marketing company. Anyone else get these? Is this really how the world of search engine marketing works?

And the real question… how much money should I ask for?

Comments (View)

May
25th

Mon

Pick One. An experiment in preference elicitation. This is amusing and it seems that it could be useful as well as entertaining. An example could be overcoming the cold-start problem in recommender systems when you have no data about a user. Identify similar users (based on this type of preference information, not just movies, for example).

Comments (View)

May
8th

Fri

NASA - Total Solar Eclipse of 2009 July 22 Total solar eclipse during SIGIR this year… unfortunately the best viewing will be at the other side of the globe. We really should schedule SIGIR around interesting celestial events.

Comments (View)

Math and Art: Differences and Similarities « Apperceptual Great post @ Apperceptual on a topic I’ve ruminated about before.

Comments (View)

Twitter Search will index the content of [linked] pages

—

Hey @Google - @Twitter To Start Indexing Links For Search

(and more here)

A somewhat shallow post at TechCrunch about Twitter search evolving, but there are some interesting bits in there. Everyone likes to compare Twitter search to Google, but of course they’re complementary. Thinking about twitter indexing content (not just tweets) is interesting. They have the advantage of a “push” indexing model, where users deliver links to them, rather than having to crawl to discover new content. As the linked-to post points out, it won’t be as complete as Google’s index, but it *might* be more fresh — at least for some portion of the web that’s interesting to Twitter users.

Most likely, Google makes heavy use of two types of data outside of the documents when ranking — links & anchor text from other documents on the web, and usage data from queries that result in clicks on the document. If Twitter indexed page content, this would the a third type of external data to use in ranking. How much information does this provide that’s not already taken into account from anchor text & query-clicks?

Comments (View)

Apr
30th

Thu

Facebook Quiz creation tasks Amazon Mechanical Turk.... evil...

AAAARRGGGHHH! Facebook has been ruined by these ridiculous quizzes (“which city are you?” “which muppet are you?” “which LOST character are you” and many, many more). Now I understand why there’s so many!

The proliferation of these quizzes is either a stroke of genius or evil… or both. Thousands and thousands of users are answering intimate questions about their likes & dislikes, and who ends up with all that data? What can they do with this data? Does anyone even realize that all answer data is being harvested?

Here’s a screen-shot for those of you without MTurk accounts:

Comments (View)

Apr
29th

Wed

TREC Blog Track 2009 Guidelines

This year, there will be a “faceted blog distillation” task looking at a few different aspects of the retrieved blogs in addition to relevance. Possible facets include: gender of the blogger, opinionated vs. factual, corporate vs. personal, detailed vs. shallow, expert vs. non-expert.

version control with Apple's Time Machine

Time machine is great. Its brain-dead simple and has saved my a** a few times. But, its not a proper version control system, and when I’m trying to re-create poorly documented experimental results from 2 years ago with code I have re-written time & time again, I really need a version control system that can handle proper diffs (and tags and branches…).

Here’s a some tools that seem quite handy. Although they don’t fill the gap between time machine and a proper version control system, they do have diffing capabilities, which is more than I’ve got now:

time machine diff script seems to work as advertised, generating colored diffs across all versions of a file in the backup volume.

And these I haven’t tried yet:

Time Tracker from Charles Soft

tms, A CVS-like command line tool. Think I’ll wait for the full-featured 1.0 before giving this a whirl, but look like it will be the most polished of the bunch when it emerges from beta.

Comments (View)

Apr
26th

Sun

Computer Program to Take on ‘Jeopardy!’ NYTimes on some efforts at IBM to tackle open QA on Jeopardy. Some quotes from Eric Nyberg here at the LTI.

Comments (View)