The future of captioning at DC

Photograph by Aly Beach

(Left to right) Computer programming analyst students Dillon Regimbal and Matthew Wierzbicki, both 20 participated in the CICan applied research showcase.

Two Durham College students have developed a way to make closed caption services quicker and cheaper.

Matthew Wierzbicki and Dillon Regimbal, both 20 and students in the computer programming analyst program, unveiled their project based in artificial intelligence (AI) at the Colleges and Institutes of Canada (CICan), ‘Accelerating Innovation Through Applied Research’ showcase on Feb. 12-13 in Ottawa. CICan acts a voice for publicly funded colleges.

“It was a great way to show Durham College has the capabilities of using AI to make real-world effects,” says Wierzbicki.

Artificial intelligence is defined as being intelligence demonstrated by machines, as opposed to intelligence shown by humans or animals.

The software idea was conceived when it was identified that the college pays a third party for captioning services.  The software could eliminate months of waiting and cut more than 90 per cent of the current price, says Amit Maraj, a professor who oversaw Durham’s AI project.

“For the college, it allows the professors to take a video and have it captioned much quicker than it currently is, which currently takes many months because it is done by a third party. So having that done in-house is better,” says Regimbal.

The projected started with automatic word generation and Long Short Term Memory (LSTM). The two students essentially trained the program on a piece of work, so it could determine sentence structures, grammar and punctuation.  They chose to train it with Shakespeare.

“We were kind of teaching it how to make words and sentences and making it learn how to talk,” says Regimbal adding, “we fed it books and had it write pages.”

The program works by taking the words people are saying, and it will train on certain words so it can “recognize the intonations and the sound files.” It will then recognize what certain words look like. From there, it will make an educated guess based on what it has heard before and output the words said.

In the future, users will be able to edit the captions after they have been created.

The project was presented at a showcase put on by CICan. At the showcase, colleges came and presented applied research projects. Policy updates and funding that would affect applied research was also discussed.

Wierzbicki and Regimbal got to meet MPPS, other college students and Bardish Chagger, Canada’s Minister of Small Business and Tourism.

“It was pretty cool meeting all of these important people. It was a fun time,” says Wierzbicki.

“Getting to talk to the students from all the other colleges and seeing how their projects are going. That was really good,” added Regimbal.

The captioning project is part of DC’s AI Hub, which works with ministry partners on research projects and internal projects for the school. The hub will be introducing workshops and seminars that teach about AI from the ground up.

“The AI Hub is sort of a term the we coined here at Durham College that oversees or the umbrella term that goes over everything AI-related that is happening in the college,” says Maraj.

According to Maraj, there are currently more than 60 students who are working or involved with AI Hub projects and around 12 researchers.

“It was kind of like an intro. We were kind of like a test to see if AI would be feasible and it seems to be successful,” says Regimbal.