i202 Fall 2009 School of Information, UC Berkeley

...is the idea that one's writing style can be reduced to vectors, and that by having knowledge of how writeprinting works, we can better imitate the style of other authors. Not that I have a particularly notable writing style (aside from my excessive use of adverbs, a habit I am trying to break), but I find the notion that someone could understand my style vectors and imitate my writing to be dangerous. So yeah, criminal justice system aside, writeprinting kinda creeps me out.

http://thedailywtf.com/Articles/Special-Delivery.aspx

XML mixup leads to trader purchasing 28000 tons of REAL coal, instead of coal futures...

Not entirely sure if this is true, but makes for an amusing story of the impact of XML screw-ups. :)

Sorry... it was a good question, but this is Craig's response:

"Dan, thanks! Can't disclose, though, any measures to detect bad stuff.

Craig" 

Hey, I'm just passing along something linked from DaringFireball.net. Blame Gruber (generally a good strategy, regardless).

http://www.cnn.com/2009/TECH/12/02/google.search.obama/index.html

The picture is no longer the top hit when you search for 'michelle obama' but is when you search for 'michelle obama ape'.

I liked this quote:
"A result that you're not looking for is not a good search result," said Rubin, the Google spokesman. "It's not a good search experience."

I highly recommend the book Logicomix (http://www.logicomix.com/en/). Yes, it's a comic book. If it helps fortify your courage to be seen reading this in public, it's gotten good notices in the New York Times Book Review.

It won't teach you how to manipulate formal logic equations, but it gives a great background to some of the big questions of the age, and might be great for some of the more technical people who will live and die by computers. Who knew that von Neumann and Turing were heavily influenced by all this?

Here's an online resource that uses NLP/writeprint to detect plagiarism.

www.plagiarism.org

They also claim to "scour" the web looking for papers to compare against...

 So there was some debate over whether translating back and forth between languages is a fair assessment of a translation system, but I definitely find it entertaining. Translation Party repeatedly translates a phrase between English and Japanese until an equilibrium has been established. For example, "May the force be with you" is eventually transformed into "One of the other five have one or two months and one single-supply one."

As some of you are probably aware, the Google Voice service will automatically transcribe your voicemail using Google's speech recognition technology - it was mentioned in one of Emily's blog posts.

Today's lecture on machine translation and language bias reminded me of this chestnut:

What do you call someone who speaks multiple languages?

Multilingual

What do you call someone who speaks two languages?

Bilingual

What do you call someone who speaks one language?

American

Learning a foreign language from translations can be a good method, but seems like some metaphors and idioms don't translate well (transcript here).

Yes, a very late-semester kind of post... .

Interesting note: Hungarian is related to almost no other European language. The most likely relative might surprise you.

This looks like fun... http://www.ling.ed.ac.uk/lec/LEC/Evolution_Experiment.html

Language creation like a big game of "telephone". How could such a quick (not over generations, and translations) evolution affect translations and organization?

Wednesday's lecture about mobile and multimedia IR got me thinking about the great This American Life episode about people who try to quantify the unquantifiable — and specifically about the two artists who tried to distill what elements would combine to make the songs that would be the most and least pleasing to the greatest number of people.

According to this article, Microsoft researchers have seen some limited success is reading subjects' minds and extracting some tagging info. That is, when exposed to images of human faces, the subjects' brains fire off in certain areas more so than when shown images of cars, non-human animals, etc. The researchers think that though this is greatly limited and specific at the same time, it shows a possible way to add tag info to images in a more rapid and "automatic" way.

For several weeks now, Coke Zero has had spooky looking ads on the net saying: "If Coke Zero has Coke's taste, is it possible someone out there has your face?" The teaser features two faces that obviously belong to different people but look roughly similar.