This mashup was created for the ClearForest Semantic Web Services Contest
and won third place.
On the back end, I created a script that parses all the news feeds from a particular news source (CNN, BBC, USA Today), keeps track of how many articles each found entity appears in, and then uses Google Images to find appropriate images for each of the entities.
An XML file records the results per category-- Person, Company, Product, City, Country-- per day.
On the front end, the current manifestation is a Google calendar "web content." This is a recently added feature to Google calendar, which lets you embed events that will show a small website or image when clicked on, right inside the calendar. This has so far been used for holiday images, moon phases, and movie releases. By including the Not'd calendar, the user can click on the little newspaper icon each day and be presented with
the Person/Company/Product/City/Country of the day, with images and articles.
The purpose of this is two-fold. First, it's a new way of browsing the news. I personally don't find myself browsing traditional news sources, but now I find myself curious to know who the person or company of the day is, and occasionally read the relevant articles to find out why they are.
The second purpose is to have an objective way of determining, e.g., the "Person of the Year," in terms of newsmaking.
Since the app is so new, I haven't been able to do any long-term calculations, but I hope to do so after I accumulate more data.
I'm also currently looking into more ways to show the data-- specifically, an image-tag cloud, with size of image corresponding to frequency.
This is not yet ready for release however.
[ add comment ] | [ 0 trackbacks ] | permalink |
( 3 / 902 )Watch the TV segment now!
Fox News LA came to USC GamePipe a while ago and interviewed fellow TAs and myself about our experience in the game program. Due to their fascination that a girl was working in the video game lab, I was put in the foreground or background of most of the shots. It was all well and good until I was forced to play Madden football. Those were the most incomprehensible 3 minutes of my life. Those who go to football games with me know I go only to check out the creative ripped-up USC shirt fashionry, and to make up inane things to shout in deep manly voices.
In the segment, you can see two games I worked on -- Firescope, shown as a burning building in Downtown LA, and Dukats, shown as a cartoony (white trash) character traveling around a building-littered valley.
For Firescope, I created the environment. The buildings are an automated simplification of GIS data, and the ground is textured with various levels of detail based off Google Maps textures. More info here
For Dukats, I worked on the game design, modeled and textured the environment, characters, props, etc., and programmed the personality engine. More info here
[ add comment ] ( 41 views ) | [ 0 trackbacks ] | permalink |
( 3 / 903 )Try out the DCT encoder/decoder now!
Our assignment for CSCI576 (Multimedia Systems Design) was to write a DCT-II decoder/encoder. For a good detailed explanation of DCT-II, read this paper. Basically, DCT-II is what's used for JPEG compression. It's a type of Fourier transform. For each color channel (R,G,B) of every 8*8 pixel block, the DCT encoder analyzes that block in terms of 64 basis functions of increasing frequencies, and finds the coefficients for the basis functions so that when they're added together, they create the block. Typically the coefficients are quantized for sending over a network, so instead of sending 0-255, it'd send 0,50,100,150,200,250, e.g. Quantization decreases the number of bits needed to send the information. The DCT decoder recieves the (quantized) coefficients, and performs an inverse DCT to figure out the intensity of each pixel.
Anyway, the best way to understand it now is to use the applet. You can try image1.rgb, image2.rgb, image3.rgb, or image4.rgb for now. The applet does the encoding and decoding of each block.
[ add comment ] | [ 0 trackbacks ] | permalink |
( 3 / 665 )Start exploring MashPlanet!
I was following the GoogleMaps Mania blog for a while thinking how cool it was all the useful map mashups coming, but also thinking how it was difficult for the casual webuser to find the mashups relevant to their interests/locale. So I created MashPlanet, which sorts all the mashups I found on the Google Maps Mania blog, Yahoo Maps! Gallery and ProgrammableWeb by locale and category.I need also to add rating and search capabilities to it, particularly rating so users could decide the most useful, e.g., housing mashup for a particular area.
[ 2 comments ] ( 2881 views ) | [ 0 trackbacks ] | permalink |
( 2.9 / 663 )Play PhotoMunchrs now!
or...
See the visual ranking results!
After I created Flicktionary and realized it could learn better rankings from users' game-playing habits (see below), I wanted to find a game that could be used to get the most visually relevant picture for any word or phrase. Flicktionary was too limited. Luckily, someone reminisced about playing NumberMunchers at a party last weekend, and I realized that would be the perfect game.
There's a more in-depth explanation of the game and visual ranking at the second link. Please play- the more you play, the better the ranking!
Skills Used: ActionScript, JavaScript, PHP
Programs Used: Flash
[ 1 comment ] ( 14 views ) | [ 0 trackbacks ] | permalink |
( 3 / 812 )Next
Categories
