The first thing we did this morning was breakfast like usual and then we went into the talks. They're really interesting and not too long so I'll cover them now.
Words and Meaning: a computer scientist's view
Lyle Ungar
This guy believes in the existence of tokens, but not words. If you're a linguist, you'll know what that means; if not, don't worry about it. I hate talking about what a word is because it's a ridiculous argument (meaning that I don't think either side has a good argument, and for the sake of my sanity I just ignore it altogether). So! Dr. Ungar has done work with Facebook and the web in general and modeling language thereof. Basically, you can use the instances of certain words all across the internet and make a statistical model showing which words are most likely to come before, after, around, or in the general vicinity of other words. By making these statistical models, and by also attributing different classes to each word, you can create a digital multidimensional space that shows which words are closer in function to other words, close in meaning, close in (insert description here) by determining how far away they are from each other in whatever dimension.
By doing a principle parts analysis and mushing these clusters of words onto a 2-dimensional plane, it's easier to see how words are related, how their relativity gives them semantic content (a principal called Distributional Similarity, a model of semantics hypothesizing that words are attributed meaning according to other words), and what words they are grammatically like. Using this model, computers that interpret language can tell, by calculating the distance between words inside the statistical model, which sentences are sarcastic or using metaphor, and which meaning of a word is intended (rose the color versus rose the object versus Rose the name). Dr. Ungar also described the process by which search engines can learn new associations and generate lists of data. I didn't explain that very well so I'll give an example: a computer scientist gives a computer a small list of word pairs to represent something that contains useful information, like movies and their directors. So he feeds in James Cameron - Avatar, M. Night Shyamalan - Every Crappy Movie Ever, etc, and the computer then finds on the web all the instances where those pairings are present, makes a statistical model of all the words that occur frequently in those contexts, and then finds those contexts in other websites to draw out the needed information for all the rest of the movies for which data is available somewhere on the internet.
He also talked about how language information was being drawn off Facebook from users who took a specific personality test to see if people with specific personality traits used certain words more or less often than other people. This was pretty funny. People with anger issues or people who are introverted, quiet, etc., tend to use words like "pokemon," "manga," "anime," and a slew of curse words, while agreeable people tend to talk about Jesus a lot and other stuff. He showed us a lot of collages showing the general trend for each demographic and I got a shot of the best one for "Less Agreeable People":
Next!
Intelligent Robotics
Daniel Lee
Usually I'm interested in intelligent computing but I wasn't really excited about this lecture until about halfway through. Dr. Lee works on the robots here at Penn, which they build to learn how to balance themselves, play robot soccer at international tournaments, and do lots of work on integrating sensory perception into machines. They make laser scanners and use cameras so their systems can see where they're going and what not to run into. The robots can recognize objects in their surroundings and learn what to avoid. Also, they can distinguish color in order to mark their goal and coordinate leg/arm movement (specifically while playing soccer).
He talked about how the human mind integrates aural information by using triangulation to tell where sounds are coming from even in the absence of a visual stimulus, both vertically and in the horizontal field. However, when the shape of the ears is changed, we lose the ability to triangulate and have to re-learn how to understand where sound is coming from in the environment. So that was pretty cool. Also, did you know that in order to tell where things are in the horizontal plane, our brain integrates the aural signal from the right and left ears by measuring the length of time it takes for the sound to come in one ear as opposed to the other and then interprets that less-than-a-millisecond length of time into a position in space? Whoah.
Also, owl ears are asymmetrical. One points up, the other points down. Pretty cool.
__________________
After the lectures today we went to lunch and I passed out for half an hour. We were scheduled for a tour of South Philadelphia that I was determined to go on. We took the subway from campus down to South St. and started walking around to look at stuff.
If you've never been to downtown Philly, I would absolutely suggest spending a day or two there. There are shops upon shops upon shops, and each building looks like it was made with entirely different architecture from the ones around it. Additionally, at least on South St., many of the buildings are adorned with artwork that looks like... well, I'll just put a couple pictures up.
Also,
nanananananananananananananannananannanaanannananannananananana Batman!
After walking around for a bit we got hungry so we headed over to the two restaurants which both claim to have created the Philly Cheesesteak; Gino's and Pat's. It's odd... they're right across from the street from each other.
The way you order Cheesesteaks is by saying "cheesesteak whiz wit." The cheesesteak part is obvious, but the "whiz" means you want cheese whiz on it (it sounds gross but it's sooooooo good) and the "wit" part means with onions. If you leave those parts off, you get it plain. If you say "wit whiz" like I accidentally did, they will ask you again if you want onions. Also, the people that work at both places are kind of ass holes. At Gino's, there is a picture of the owner with a sign that says "I want my country back. If you come here, order in English!" The premise of this is absolutely ridiculous given that his family is from Italy, but whatever. At Pat's, they just really hate you and wish you would move faster. Also, if you're a girl, they will call you "babe" about 4 times before you can make it to the end of the line. I suppose the trade off for all this is that you get your cheesesteak really, really fast. Like, as soon as you hand them your money, you have your sandwich. 
I had two; one from both places, just to make sure which one was better. It's Pat's. Definitely Pat's. But I wouldn't advise eating two cheesesteaks in a row; it's a really, really bad idea.
After the cheesesteaks we split up into like three different groups; one group went back to the subway, one group went another direction, and the rest of us went down to the hipster part of town. There's a whole street that used to be home to all the Mexican immigrants, but it has apparently, over time, evolved into a miniature hipster paradise. There are plenty of vintage shops, old clothing stores, a farmer's market, etc. If you go shopping in South Philly, though, you're going to want to bring a lot of cash. Many of the places don't take cards or checks, so keep that in mind. I made an on-the-spot decision to buy an awesome shirt to go out in tonight for $20 at a local men's clothing store that was full of fancy everything and ridiculous shoes.
I created a group on Facebook for all the people here at the conference by stalking them all using their e-mail addresses, but the method was effective. We've all decided to go to college night at the gay bar, Woody's, down in middle-south Philly. I'm really excited to go out in my new shirt with all these awesome people. More later!
No comments:
Post a Comment