Login is restricted to DCN Publisher Members. If you are a DCN Member and don't have an account, register here.

Digital Content Next


Research / Insights on current and emerging industry topics

A Covid-era study reveals best practices for automated reporting

August 17, 2021 | By Rande Price, Research VP – DCN

Though automated journalism is not new, the pandemic — characterized by a slew of data driven stories — accelerated adoption for some newsrooms. By in large, automated approaches use a natural language generation (NLG) application to automatically transform data into human-readable text. It has long been used for weather and sports, as well as financial market updates. However, it has been making significant in-roads at a wide range of newsrooms and other coverage areas over the past decade.

Early adopters included The Los Angeles Times, which used NLG to for earthquake alerts and to report on homicides, and the Associated press, which used it for corporate earnings. Today, numerous media organizations are adopting automated journalism in their newsrooms and a wide range of storytelling practices.

In a new report, Samuel Danzon-Chambaud, a Knight News Innovation fellow at the Tow Center for Digital Journalism, examines how automated news has been used to cover the pandemic. Covering COVID-19 with automated news outlines use cases at nine news outlets in eight countries to reveal both opportunities and stumbling blocks in adopting automated storytelling processes.

The media organizations include:

  • Bayerischer Rundfunk (Bavaria’s public service broadcaster, Germany);
  • Bloomberg News (news agency, United States);
  • Canadian Press (news agency, Canada);
  • Helsingin Sanomat (newspaper, Finland);
  • NTB (news agency, Norway);
  • Omni (news service, Sweden);
  • RADAR (news agency, United Kingdom);
  • Tamedia (media group, Switzerland); and
  • The Times (newspaper, United Kingdom).

Rethinking the process

Using automated storytelling tools requires newsrooms to rethink their process from data sources to workflow and operations. For starters, it’s important to ask the following questions:

  • How will the automated content serve readers?
  • What metrics are needed?
  • Who are the data sources?
  • How often will you update the data?
  • How will you upload the data into a usable information?
  • Will your CMS allow easy data integration?
  • How does the automation process fit into the current workflow?

In terms of the pandemic, automated news coverage offered timely updates on the spread of the virus. News outlets were able to provide audiences with dashboards and newsletters as well as automated or semi-automated graphics. Publishers also created new news products such as an automated-live blogging platform.

Case studies show challenges met in real-time

Bayerischer Rundfunk (BR), Bavaria’s public service broadcaster questioned which metrics to track. They started by tracking absolute increase in Covid cases and total numbers of cases. However, early in the process they realized the data did not allow them to track the evolution of the virus. They quickly pivoted and switched their focus to access more detailed data points such as doubling times, reproductive numbers, and 7-day incidence rates. This allowed for more robust reporting and updates on the virus.

For some news outlets, CMS issues can prevent full-on automation. At The Times, a newspaper in the UK, the process could not be completely automated. After the text was generated and edited, it had to be copied and pasted it into the newspaper’s CMS. Unfortunately, the numbers were often outdated by the time the story was ready for publication. The data and interactive team were called upon to change their workflow until the automation process could be integrated into the CMS. The new workflow allowed journalists to update the data in their articles without an editorial review, unless the story’s lead changed. 

Similarly, the automated tool built at the Canadian Press (CP) asked journalists in the field in different regions to input different Covid statistics into a master spreadsheet. The spreadsheet was exported into a JavaScript Object Notation (JSON) file which stores the information in a simple data structure. The data is then transformed into automated stories and graphics but on separate webpage. Again, the need for  copy and paste was necessary to place CP’s automated content in their CMS. CP will soon have a CMS that will generate and integrate an automated story using a simple Slack command.

Using automated news to develop new workflows

Bloomberg News, a news outlet in the U.S., uses their automated news process to extract relevant information on companies’ financial statements and reports. The automated system is connected to an AI-powered system to analyze the data through “knowledge graphs.” These graphs combine different data silos and illustrates the relationship between them.

Bloomberg’s process includes machine-learning elements in the analysis of the knowledge graphs. However, it’s also important to point out the human effort is needed in writing scripts in advance to prepare for the potential scenario in the analysis. Bloomberg News was able to include their business insight in their Covid-19 analysis to assess the economic impacts of the pandemic. Their Covid-19 analysis included assessment of metro riders to flight reservations, restaurant bookings, and supermarket visits.

Automated news invites new technology — and opportunity —into the newsroom. Increasingly, newsrooms are thinking of new ways to incorporate automated practices into their product development roadmap. Next steps include ideas such as an alert system to inform teams when new data is being released and cloud storage services to keep track of historical data.

The more that media practitioners engage with automated news systems, the better their ability to manage the process and foresee potential disconnects. Importantly, keeping reporters involved in this process will ensure that journalistic standards are encoded into the automated content creation process.

Print Friendly and PDF