The French Revolution, which began in 1789, was the bloody scene setter for a myriad of European political upheavals. Now, machine learning is shedding light into how linguistics played a role in the discussion of democratic ideals and formation of the new government.
A team of researchers, including Carnegie Mellon University assistant professor Simon DeDeo, used machine learning to analyze more than 40,000 digitized transcripts from the first two years of debates of the first makeshift French parliament, during the beginning of the revolution.
"The first thing that really came out and surprised us was that you can distinguish left and right not by what they're saying, but by how they're saying it," DeDeo said.
The study found liberal revolutionaries were more likely to use novel turns of phrase to talk about new ideas, and they also did more discussion derailing.
"So you have these really charismatic people who are essentially the rudders of revolution," he said. "They're steering this conversation about how to run France into directions that nobody's ever seen."
Conservatives, on the other hand, used more traditional language and speech patterns, and they tended to keep conversations on track. DeDeo noted that the situation for conservatives in France became increasingly hostile as the revolution raged on, and many were forced to flee.
DeDeo said the study confirms what many historians have thought about the roles of different revolutionaries -- which is a good thing.
"Because now what we have is a computational, mathematical tool that we can apply to a lot of other different political systems," he said.
DeDeo said the research team plans to use this machine learning to analyze the democratic transitions of other countries, such as Serbia, that are far less understood.
