Hacking away at the Next Big Thing

The Next Things in tech tools are born in rooms such as this one at the American Can Company in Canton, amid laptops, Starbucks cups and Aquafina bottles, amid 20- and 30-somethings working in groups or alone to the sounds of keystrokes and soft conversation.

This is the Baltimore Hackathon, the first ever, in which computer cognoscenti step out of their electronic communication worlds and gather to share ideas and physical space, pitting projects against each other in competition for cash prizes. From Friday through Sunday, about 50 participants — almost all men in their 30s or under — are working long days and into the night in hopes of finishing their software or hardware venture by Sunday afternoon.

"We wanted to have an outlet for people who want to build stuff, create that space," said Matt Forr, one of four organizers of the event, which was modeled on gatherings that have been held in cities around the country for a couple of years now. Working at home alone is one thing, but the gathering fosters connections that even the best social networking cannot quite match.

"The community access," said Jonathan Julian, another one of the organizers. "You get to meet new people."

Tech mavens from Baltimore and elsewhere in the Mid-Atlantic convened on the third floor of the Can Company building Friday evening, spread their gear out on long folding tables in several rooms at the offices of Emerging Technology Centers, one of 15 event sponsors, and went to it. Some brought only their laptops — Mac is the clear preference in this crowd — some brought boxes containing circuit boards, tiny motors, and a gizmo that can create items in plastic from instructions conveyed from a computer.

Zach Waugh and his work partner, Adam Bachman, were trying to make a computer application that would enable users to generate and share lists of things to do or buy while surfing the Web. Amy Hurst and Marty McGuire were working on a combination software/hardware project they hoped would be able to create an array of products, such as rings and tool handles, custom-sized to a particular hand. McGuire, who works for a New York-based robotics company, brought the machine that makes items in plastic based on specifications fed to it via computer.

The projects are to be judged at 4 p.m. Sunday, with prizes of $150 for the best individual project and $350 for the best group project.

Matt Miller, Jake Wells and Michael Smith came down from Philadelphia to work on a project that takes another step in language recognition and analysis. Programs already exist that can take a chunk of written text and display it on the screen in what is called a "word cloud," or "tag cloud," meaning a cluster that shows the words in sizes proportionate to how often they were used in that cluster of text.

The effect is to produce a snapshot of the prevalent concepts in a block of text. Smith showed how it works using the front page of The Baltimore Sun website, which on Saturday afternoon generated a word cloud with "Ravens" as the largest word in the cluster.

"Our idea was to do this with conversation," said Miller, who along with Wells and Smith is a student in the Master of Industrial Design program at The University of the Arts in Philadelphia.

Their work could conceivably create an application to record, say, a business meeting and present a visual summary of the most prevalent concepts.

"It could point out inefficient meetings if the most common words are way off topic," Miller said.

They arrived Friday evening thinking they'd be working with CMU Sphinx, an open-source, or free, voice-recognition software developed at Carnegie Mellon University. But at the Hackathon, they bumped into Mark Headd, who came down from Delaware where he works for Tropo, lead sponsor of the Baltimore event. Tropo acts as a host site for all manner of Web-based communications, including Twitter and Instant Messaging, and also offers voice recognition programs for various uses.

Headd said Tropo provides tools that can make life easier for application developers, saving time that might be spent coding or entering data to get the tool to perform a particular task.

Indeed, Miller and his team found the Tropo voice recognition easier to use than Sphinx, although it wasn't perfect. Miller discovered that when he leaned into his MacBook to perform a test.

"Testing 1234. This is a test of the accuracy of Tropo," Miller said. In seconds the display came back: "Testing 1238. This is a test of the accuracy of troll 12 …"

All right, it was only Saturday afternoon. Still 24 hours to go.


Copyright © 2020, The Baltimore Sun, a Baltimore Sun Media Group publication | Place an Ad