When the 2014 CAR conference open call for lightning talk proposals was sent out, the session became the one I was most looking forward to at NICAR 2014. Moderated by Derek Willis of the New York Times, each presentation was entertaining, informative and (as advertised), fast. This year they were sponsored by the Knight Foundation, positioned in a primo Friday night slot on the schedule with no other sessions scheduled at the same time. Which meant the entire conference was in attendance, crammed into three of the hotel's ballrooms, with every single seat occupied and people sitting on the floors and standing in the back.
The first NICAR lightning talks were in a small conference room, maybe 40 people showed, 10 were speakers. Now? pic.twitter.com/XJBFAAyzZx— Matt Waite (@mattwaite) February 28, 2014
1. Refactoring; or Why Your Code Sucks and How to Fix It
Christopher Groskopf, a developer at NPR, started off the Lightning Talks explaining the importance of refactoring your code. By improving the quality of code without adding more features you can make it more readable and efficient. Following "code smells" like duplicated code, long functions (like longcat), and inconsistent style helps build better habits the next time you code.
2. A Few of My Favorite Wee Things
Lena Groeger, a news apps developer at ProPublica, then shared a few of her "favorite wee things," small visualizations that are easy to create but convey lots of meaning. Small multiples of images show variations over time and small text makes differences stand out. Mini maps provide context as a reader scrolls through a story. Icons are useful way to add information inline with the text. Groeger referenced The Noun Project and StateFace as tools to consider for finding these small images.
Anthony Pesce, a journalist at The LA Times, explained how he used Natural Language Processing (NLP) to turn the paper's recipe archive into a structured database. Using the Python library *nltk* he was able to train the program to classify words and phrases as either ingredients, steps, or instructions.
In another more computer science heavy talk, Chase Davis, explained five algorithms and computer science concepts in five minutes. Explanations and demos of how to use vectorization, naïve-bayes classifiers, iterative algorithms, vantage point trees, and latent dirochlet allocation are all available on GitHub. Watch the video.
Medill School alumna, and former contributor to Knight Lab projects, Katie Park talked about learning from bad data visualizations. In between laughing at some really hilarious examples, Park cautioned against always using charts, stressed being consistent with scale and color, and emphasized double and triple checking your data. Watch the video.
Steven Rich talked made an argument that math can find an investigator or journalists answers that can't be found by any other way. Rich talked about how he used calculus to fact check the government's estimate of how much toxic chemicals had spilled into the Elk River in West Virginia. Rich concluded by emphasizing that journalists can and should use math.
7. Detecting What Isn’t There
Sisi Wei, a Medill School alumna and developer at ProPublica, talked about how ProPublica reverse engineered the Sina Weibo API to discover what posts were being censored. She explained the challenges and pitfalls that are important to consider when trying to be discrete with your online investigations.
In the fastest talk of the hour Jeremy Bowers, another developer at NPR, gave his in, well, the five minute limit. Speaking at around 350 words a minute (you can read the full transcript), Bower's talk was definitely entertaining and mildly informative about the technical infrastructure that powers the Internet. GitHub repo | Video
The Lab's senior student own fellow, Tyler Fisher, then talked about "How to Raise an Army" of news nerds. Fisher talked about the steps the Lab has been taking to grow more future digital and data journalists. He explained how the Open Lab Hours and brown bag lunches and Open Lab Hours have been effective for growing a community around students learning programming for journalism skills. He talked about the inspiration and impact of Learn.KnightLab.com. We were all quite proud.
10. You Must Learn
Ben Welsh was granted the rest of the available time for his talk about how, in some sense, using data and computers for reporting is nothing new. Journalism developers should instead be proud of their history and work to make data journalism even better. He emphasized the importance of understanding the history of the craft in order to more effectively invent journalism's future. Watch the video.
Each talk was informative and entertaining and provided a great break from the more traditional (but still fantastic) sessions at the conference.
Every year New York Times' Chrys Wu curates 'Slides, Links & Tutorials' from the annual CAR conference: 2014, 2013, 2012 and 2011. They are a fantastic resource for all journalists and investigators of all skill levels.