Students Present Innovative Projects in Journalism and Technology
One of the projects, TED 2.0, automatically generates scripts for TED talks
Coming soon to a computer near you: a program that generates a script for a TED talk, a tool that finds sources for journalists, and an application that builds an itinerary for short vacations.
These are just three of the projects that emerged from Northwestern’s Collaborative Innovation in Journalism and Technology course. A joint endeavor of McCormick, Medill School of Journalism, Media, Integrated Marketing Communications, and the Knight Lab, the class requires computer science and journalism students to work together to design and implement innovative solutions in technology and media. Teams presented their final projects on Wednesday, June 5 in the McCormick Tribune Center Forum.
Rich Gordon, professor of journalism and director of digital innovation at the Knight Lab, called Ted 2.0 “the most unusual group” to emerge from the course. The team created a system that automatically generates the script for a TED talk by assembling information about a selected topic, such as fashion, health, or comedy.
The team—comprising McCormick students Divir Gupta and Arundhati Jaswal, Amanda Anumba from Weinberg, and Yang Zhang from Kellogg—started by asking what makes a great TED talk. After analyzing numerous speeches, the team decided that all great talks had emotional appeal, personal anecdotes, and clear structure. They designed an application to generate scripts with these three attributes by taking text from the opinion website debate.org. By fusing different opinions about a particular topic along with comments on the site, it organizes a script. The user can select an avatar to deliver the speech aloud.
The system uses supervised learning, a machine-learning task that uses known datasets to make predictions. Through this technique, the team will continue to teach the program what makes a good talk, and it will get better at generating more complex and emotional TED talks.
“Through machine learning, it will eventually speak on a college level,” Anumba said. “Right now it speaks more like a grade-schooler or kindergartener.”
Another team built a program called Sourcerous, which finds sources for journalists. “When pursuing a story, sometimes the journalist doesn’t know, off-hand, who to go to right away,” said Curtis Sprung, a team member and journalism student.
Sprung—along with team members and McCormick students Thomas Morreale, Jin Sun, and Xiaofeng Zhu—created a program that reduces the time locating sources so journalists can focus on reporting. The user types in the name of a company or institution and within seconds Sourcerous displays a comprehensive list of names that have appeared in articles with the topic in the past. The team views this as a “jumping-off point” to help journalists get started.
Eight teams presented their prototypes to the audience. The other presenting teams were:
- The Weekender: An app that builds an itinerary for a short visit to an unfamiliar city.
- Tweet Talk: A browser extension that automatically finds relevant tweets from experts about a given news story.
- Trendable: A system that helps assemble and analyze time series data.
- Print Share: Social media sharing for articles in print publications.
- VizAnalytics: A tool enabling online publishers to benchmark their performance based on web analytics.
- Qwotd: A browser extension that displays the Twitter conversation about any page a user is viewing.
- WikiNow: The news “front page” for what’s trending today on Wikipedia.