Can-PolNews: A Multi-Platform Dataset of Political Discourse in Canada

Résumé analytique

For many societies, social media has become the primary venue for encountering and engaging with political discourse, but publicly available datasets are often limited to a single platform even though ideas and conversations move fluidly between different platforms. This paper presents Can-PolNews, a multi-platform social media dataset focused on political discourse in Canada, spanning January 1, 2023 to January 1, 2025. The dataset contains all content posted to social media by Canadian news media (national, provincial, and local) and Canadian politicians (federal and provincial) for 1,852 unique accounts across four major platforms popular among Canadians: Instagram, X/Twitter, TikTok, and YouTube. Politicians are labeled by their political party affiliations and provinces, facilitating comparative analysis of regional political trends and ideological affinities.

By covering a two-year time frame, the dataset contains more than 5 million posts with a normalized schema across the four platforms, enabling researchers to analyze patterns and trends of digital political engagement and the interplay of news and political elites on social media in an established democracy. It also includes bilingual (English and French) content, reflecting Canada's bilingual and multicultural context, making it the largest publicly available dataset of Canadian political social media activity.

Points saillants

• The dataset contains over 5 million posts from 1,852 unique accounts across Instagram, X/Twitter, TikTok, and YouTube, spanning January 2023 to January 2025.
• Coverage includes 1,071 politicians and 781 news outlets, annotated with party affiliation, province, and electoral riding.
• X/Twitter has the highest platform penetration for both politicians (92.6%) and news outlets (92.3%); TikTok has the lowest (6.3% politicians, 8.1% news outlets).
• The dataset is bilingual (English and French) with normalized engagement metrics harmonized across all four platforms.
• This is the first two-year, multi-platform, multi-entity dataset of Canadian political discourse, enabling longitudinal analysis across major Canadian elections.