News / Toronto

‘Hatebase’ aims to prevent genocide by tracking hate speech

A Toronto-based NGO is tracking the use of hate speech around the world, hoping to help stop genocide before it starts.

The Sentinel Project recently launched Hatebase, a wiki-style website that asks the public to contribute to a multilingual dictionary of hateful words and log incidents where those words are used. The Hatebase API allows users—potentially governments and other NGOs—to map out and chart localized spikes and trends in hate speech in real time.

Spikes in hate speech, when combined with other knowledge about tensions in a region, can be an early warning sign of violence, or of violence escalating to genocide, said Timothy Quinn, an advisor with The Sentinel Project.

“When we start layering in hate speech that gives us an indication—when it’s spiking over time—that tensions are rising,” he said.

For example, in Rwanda in the ‘90s inyenzi, meaning “cockroach,” was a derogatory term for Tutsi’s that escalated in use before the genocide, with radio announcers ultimately calling on people to exterminate the cockroach.

While the Sentinel Project focuses on preventing genocide, the upsurge in hateful language can also indicate other violence that doesn’t amount to genocide.

“God knows, there’s a lot that can happen before then that certainly warrants monitoring and intervention,” said Quinn.

Defining hate speech isn’t easy.

“Hate speech is really defined more by the intent than the word itself,” said Quinn. “There are some words that might have a benign meaning some contexts, but in other contexts have a very malignant intent.”

For example, the use of “cockroach” in Rwanda in the ‘90s could be attributed to a cockroach infestation, instead of a brewing genocide, without a deeper knowledge of the region.

The Hatebase dictionary is growing as more and more people with important knowledge about the world are beginning to use it, said Quinn.

“The spread of our users is global,” he said. “We’re starting to get a lot of very localized slurs, which we didn’t have before.”

 This map of Iranian hotspots comes from a screenshot of ThreatWiki, The Sentinel Project’s internal crisis monitoring software. Organizations such as the Sentinel Project that monitor volatile regions for signs of conflict can use Hatebase to add information about hate speech spikes to their monitoring systems.

This map of Iranian hotspots comes from a screenshot of ThreatWiki, The Sentinel Project’s internal crisis monitoring software. Organizations such as the Sentinel Project that monitor volatile regions for signs of conflict can use Hatebase to add information about hate speech spikes to their monitoring systems.

Politicizing words

Relying on user-generated content isn’t without its problems, said Quinn. Hatebase has attracted some people political agendas who have started fighting over words and definitions that likely have very little value for organizations concerned about genocide.

“I’ve had a couple of Nazis that are pissed off at me. I’ve got a group of anti-feminists that are pissed off that we removed a word we didn’t think was hate speech,” said Quinn.

For example, some recent additions to the Hatebase wiki are academic language commonly used in women’s studies, like “patriarchy,” “cis male” and “white privilege”—and if the words were a sign of genocide, university campuses would be very scary places.

Fortunately, most users Hatebase can distinguish what is relevant and what isn’t and wouldn't bother with anything that doesn't relate to an area of concern, Quinn said.

“Some of it is meaningful and some of it is not. Funnily enough, the stuff of dubious value is the stuff people fight about the most,” he said.

More on Metronews.ca