Newsroom Automation at the Canadian Press: Doing More with Less
The coronavirus pandemic has accelerated efforts by news organisations worldwide to ponder a seemingly intractable problem: how can they produce more content with fewer resources? For many newsrooms there are only partial solutions — such as furloughing workers or switching from print to digital. But for the Canadian Press (CP) and the news agency’s Digital Data Desk (D3), new newsroom technology and automated digital publishing solutions are addressing this issue head on.
While CP’s data desk has been writing algorithms to produce targeted stories since 2017, the pandemic — along with a planned CMS upgrade to Superdesk — has been an opportunity for the desk to play a stronger role in shaping the news agency’s trajectory.
“Our goal at CP is to switch from people writing tedious copy — like market reports and local sports reports — to having computers write those stories, which in turn will free up reporters to focus on other tasks,” says Lucas Timmons, the Digital Data Desk’s chief editor. “We’re essentially employing journalists to write code, and we’re trying to build a desk based on that idea.”
One early example of this effort is the Labour Force Survey bot, which uses an API to access monthly labour data from Statistics Canada to automatically generate stories. Every month, the Canadian government releases figures on the economy — such as jobs added and local and national unemployment rates. Using this monthly data, the bot produces 133 stories, including 129 stories with a regional focus, along with national and provincial briefs, industry breakdowns, unemployment stories, and “quick hits” for all 65 of Canada’s economic regions.
Another AI-focused project underway emerged from the agency’s coronavirus coverage. When the pandemic started, CP began a daily “by the numbers” feature highlighting cases and deaths from across the country. Using Google Sheets, Timmons created a tool to extract provincial-level data and automatically turn it into stories and charts, which in turn feeds a series of newsletters that the agency puts out.
At the moment, the data desk’s story tools are being built for the future; only a handful of the AI-generated copy is being distributed to publishers, due to limitations with CP’s current CMS. But that should change later this year, when CP migrates to Superdesk, Sourcefabric’s open-source headless CMS for news organisations. While CP chose Superdesk to improve workflows and publishing output from the news agency’s traditional desks — like business, politics, and sports — the data desk will also benefit greatly. “When we have Superdesk in place, we can build an API to access these tools,” says Timmons. “Then, with the click of a button, content will be ready to go. We could even schedule charts and stories to be automatically published multiple times per day, or even every hour.”
The data desk, which recently received more than a year’s worth of funding from the Google News Initiative, was created to strengthen access to data and to enable “the use of AI to create content from that data,” according to the GNI. As Timmons puts it, the desk is both a resource for in-house journalists and a playground for news innovation.
Other data desk products have included a web scraper to help CP’s arts writer pore through grants posted online; a watermark to make it easier for reporters and editors to apply CP branding to distributed photos; and an election research bot that brought data from official Canadian government sources into a Slack channel for easy access by journalists and researchers. The election bot was so successful, in fact, that CP made it available as a service to clients.
In the coming months, Timmons’ team plans to build additional SaaS tools, including a real estate bot. One key feature of this tool will be an AI story generator, which will give prospective homebuyers a text story about the neighbourhood they are exploring. The best part is that the content, which can be used by real estate agents as marketing material, will require no newsroom resources to produce. “We think that’s something that would be really useful for prospective homebuyers and be a product that we could eventually sell to realtors,” Timmons says.
While the data desk is working to create new sources of revenue for Canada’s national news agency, the ultimate goal is to make life easier for CP’s journalists, Timmons says. In the past, churning out more copy to meet business needs meant more man hours — and lots of overworked, stressed-out journalists. Now, it often means little more than writing a new line of code.
“That’s sort of the whole ethos behind our efforts at the D3: Let’s let computers do what computers do really well and save the humans to do the things that humans do really well,” Timmons says.
He adds: “This is not about taking away jobs from people. Rather, it’s about using a newsroom’s resources in the smartest way possible.”