TORONTO -- A bot analyzed more than 350,000 tweets sent to Canadian political candidates during the first week of the federal election campaign. It found that more than 20 per cent of them were considered toxic, with nearly 10 per cent containing threats of violence and other aggressive language.

The Samara Centre for Democracy and Areto Labs teamed up to create a bot that could analyze tweets that mention political party leaders and incumbent candidates during the 2021 campaign. The bot, SAM, uses machine learning to analyze tweets on seven point toxicity attributes. The attributes it tracks are toxicity, severe toxicity, identity attacks, insults, profanity, threats and sexually explicit nature. Severe toxicity is defined as very hateful, aggressive or disrespectful comments.

For Lana Cuthbertson, the CEO and co-founder of Areto Labs, the findings from the election campaign so far are not surprising.

"We've seen similar trends so far, not just in Canada but in some of the other countries we've studied," she told CTV News Channel on Sunday. The team behind SAM has also tracked tweets directed at politicians during elections in the United States and New Zealand.

Women and racialized candidates are frequently on the receiving end of toxic tweets, she added.

"We tend to see not just more volume of toxicity but also attacks related to their identity, for example, or specifically gendered attacks," she said.

An example provided by SAM is that of the Liberal incumbent candidates, women were five times more likely to receive toxic tweets than men of the same party.

As for those receiving the most toxicity, they are well-known candidates like Liberal leader Justin Trudeau, who has so far received the highest proportion of toxic tweets analyzed, according to SAM's data.

"Those politicians with more national reputation or presence tend to see more toxicity and more targeting mostly because more people are aware of who they are and there's a bit of a celebrity factor there," said Cuthbertson.