Innovations In News

The result is sort of like Mad Libs on steroids. Once readers select a city and an organization (or let the computer choose them randomly), the widget constructs a skeleton of a plot and randomly fills in the details from a big list of possibilities we compiled. For example, the superlative heroine described in the second paragraph-a staple of all Brown novels-might be "coltish and enigmatic" or "lithe and charming" or "freckled and brilliant." The mysterious symbol might be imprinted on a ring or etched into the floor.

The possible words and phrases are stored in an XML document called by the Flash, so it's easy to add more at any point. With more than 25 variables and hundreds of possible words and phrases, it is extremely unlikely that you will ever get the same story twice, even for the same city and cult. It may start to feel like the same old, same old after a while, but then again, so do the books themselves.

By Chris Wilson

Developing this map was a leap of faith. While the Bureau of Labor Statistics is a treasure trove of economic and employment data, the files are far too granular to eyeball for a sense of the big picture. The county-by-county employment data for this map came as a 77MB file that was far too large to stick into Excel and sort by year or location. So it wasn't until three-quarters of the way through the development of this map that I first got a real look at the data.

The story here speaks for itself; in the past year, virtually every part of America has seen a catastrophic decline in employment. While it's a bit morbid to marvel over, it is interesting to watch the red circles cascade through the country, striking first in areas like Michigan and spreading from there. Normally, I think that a good editorial multimedia project should be accompanied by a short narrative that guides readers through the data to get them started. In this case, however, the storyline was all too apparent without my gilded prose.

This interactive video presentation uses three new technologies developed by the Product Development team from The Washington Post. It will be used for special projects and regular video shows like the upcoming onBeing.

Fluid: A 3-D media navigation tool that allows for hierarchies and multiple types of dusplays. Built on Flash Actionscript 3 and using Papervision 3D, Fluid can display numerous videos in 3D space allowing users to navigate through them without scrolling.

Web-Com: A threaded and user-moderated comment system that provides a visual representation of the most valuable comments and conversations among users. The comments are cataloged in a database and displayed via Flash. The most liked comments become larger and conversations (replies) gravitate to the center. Users who like a more structured view can switch modes to see them in a list.

Omni Viewer:The Flash video player can display lives broadcast and progressive streaming at any size and in full screen. The new controls provide an intuitive interface t also allows users to quickly move through the video without buffering delays.

All three tools have a Django administration tool that provides editors with the ability to build and edit special media packages and to choose which views or technologies to use.

Currently, we are serving the videos via Brightcove but the player allows for live RMTP streams, mp3 or any flv file.

The feature was developed by the Product Development team known as the Web Ninajs. Jesse Foltz developed the code-intensive Flash interface and Lee Trout developed the behind-the-scenes databases, scripts and administration tools.

By: Chris Wilson | Slate.com

A week before the much anticipated movie Watchmen came out, my colleague Chadwick Matlin and I were wondering whether we could predict how well it would do at the box office based on the lukewarm reviews it received. The movie was virtually certain to be the highest grossing movie that weekend, given the tremendous hype and lack of credible competitors debuting at the same time, but our theory was that it would take a major hit in sales in subsequent weeks as word of how bad it was got around. (I later made the mistake of seeing the movie, and I'm inclined to agree.)

To test this theory, we decided to gather box office info from lots of recent movies and measure the percentage of their total gross revenue they made in the first weekend. The theory was that well reviewed movies would have a low percentage, since people continued attending well after the opening weekend; overhyped bad movies, meanwhile, would quickly fizzle. To help visualize the data, I built this tool in flash to plot every movie we looked at on two axes: Percent of profits from opening weekend (vertical) and average review score on Metacritic.com (horizontal). If our theory was correct, most dots would group along the diagonal line from the top left to bottom right, a basic inverse correlation. I also colored the dots by genre, in case that was important.

At first, the dots looked a big jumbled that betrayed no correlation at all. But since there appeared to be some more ordered pattern among just the action movies, I added a feature that allowed users to toggle genres on and off using the check boxes at the top. Lo and behold, when you hid every genre except action, a rather distinct correlation emerged in just the direction we predicted.

In the course of collecting all the data for this tool, we ended up with a lot of raw numbers-total revenue, revenue per theater, and so forth-that didn't make it into the original graph. But one of my favorite aspects of data-driven journalism is that, when there's too much information to process, you can build a simple way for readers to parse through the numbers on their own and discover their own conclusions. So I added the dropdown menu at the top left, which let's users choose which dataset they want to see represented on the y-axis. (Because this was developed in Actionscript, I could make use of the "tween" function that gives you that elegant transition from one dataset to the next.) We didn't find any strong trends in the other datasets, but that doesn't mean someone else will not.