Editor’s note: This special edition of toolbox looks across multiple newsrooms and their innovative experiments. Previous editions of toolboxes, focused on individual publications, can be found here and here. Want to submit a newsroom to be featured in the series? Writer to firstname.lastname@example.org
Some of the best knowledge sharing in journalism comes from newsrooms that push the boundaries of new technologies. These are not exclusively big newsrooms with big budgets like The Guardian or the BBC. You can find media innovation in many small newsrooms.
But if you want case studies and experimental news products, a few outlets are worth paying extra attention to. They regularly share their explorations with details that can guide your own experiments.
Let me share some of those that caught my attention in recent months. Although there are many blogs I could have chosen from, for this column I will be looking at recent experiments by The Guardian, BBC, The Financial Times, The New York Times and The Wall Street Journal.
Last December, the JournalismAI project at Polis (the London School of Economics think tank) held a week-long JournalismAI Festival. BBC News Labs was among the case studies (you can find the session recordings here). They explained how they use automation to convert old stories into new formats or use technology to enhance existing articles.
The News Labs team showcased a couple of their experiments. One used Google’s AI models (Pegasus & Bert) to summarize an article into bullet points. You can then use them in the beginning of the article. Another example was practical personalization for news where journalists can take an audio clip and use a tool code-named CuPid to create other formats.
But there was one experiment I thought would be super useful as a tool for any newsroom. Simply called the Graphical Story Editor, it lets journalists quickly create story formats for social media from text.
The tool is a graphical editor into which you feed a story. It has a catalogue of images and some pre-made templates. It then takes the content and creates a whole set of stories for you. Of course, journalists can edit the stories or build new ones on their own.
Still a prototype, the team is thinking about integrating the tool into their CMS. “We are exploring the idea of integrating the tool with the new content management system currently in development at the BBC and even publishing the stories in the BBC News mobile app,” the team noted.
More from The Fix: 5 innovative news formats to try out in 2021 – and tools to implement them
I like to read Inside the Guardian, The Guardian‘s blog on changes and updates, as well as its Engineering blog. Both share interesting stories and lessons learned, or show off things they did that might be useful for others.
You might have caught the blog where The Guardian announced it will make it more transparent off-platform (mainly social media) when a story was published. Since then I have seen this initiative spread to other media outlets, too.
Last November, the blog introduced Typerighter. This tool takes information from the paper’s style guide and flags to journalists when they use terms incorrectly. Here is how the blog described its success:
It proved immensely useful to the newsdesk and spurred us on to develop the tool further. A team of journalists and developers from across the organisation spent the next few months building our entire style guide into the Typerighter tool we have today.
It has been built into our existing in-house editing software – Composer. We can switch it on from our text-editing tools – and it works a bit like a Guardian version of the well-known editing tool Grammarly.
It acts as an aid to journalists and subeditors, quietly helping out in the background, but never getting in the way of focusing on maintaining the quality of the writing. As for subeditors, we have more time for the other aspects of our job – the headlines, pictures and compelling standfirsts.
Of course, the vast majority of our readers won’t ever see Typerighter doing its thing; what they will hopefully see is what they want to read produced more quickly, accurately – and presented with the kind of care they have come to value.
More from The Fix: Too many news media are terrible at communicating their value proposition
The Journal’s Digital Experience & Strategy team (DXS) run an excellent blog on Medium (+ they also have a podcast). There they share how they think about product, design, technology and audience strategy.
In a recent blog they focused on retention strategies for subscribers. As many media outlets, WSJ has also seen a surge in numbers and the team wanted to get new and less-engaged members to come back more often to the site (they find it improves subscriber retention).
First up: a daily email offering a convenient link to the PDF version of WSJ’s print edition. The pandemic cut into single-copy retail sales and home delivery, so we wanted to ensure this experience wasn’t an insider “secret,” but easily available.
‘Next, smaller changes, which showcased habit-forming features and delivered tangible results. A prominent “View Watchlist” link on our desktop homepage led to a 90% increase in unique visitors to that feature.
A new “Podcasts” link in our navigation drove a 16% increase in podcast plays in its first month. Extending “related newsletter” promotions to Google AMP stories ensured we invited mobile search audiences to sample more of the Journal’s reporting.‘
In the blog, they share they hit a big success with the introduction of the ‘Listen to this article’ feature. This provides an automatically generated, text-to-speech audio version of every story on WSJ.com. As the team noted, it proved to be more habit-forming than their popular crossword puzzle.
More from The Fix: Ukrayinska Pravda: Bringing innovation to the membership model
The Product & Technology blog by the Financial Times is another must-read. It features guides like this one or results of their experiments, such as this blog on how the paper used the bubbles stories format from social media to drive engagement with myFT (an FT feature which enables readers to select topics of interest to follow).
A more recent experiment described on the blog focused on predicting trending topics.
‘Understanding the preferences of Financial Times readers is crucial for improving user experience and maintaining engagement with our products. Having accurate indicators showing which area is increasingly important can augment journalists’ work, by helping them to focus on topics of interest.
Trending topics prediction is a data science model built using machine learning and time-series analysis. We define article topics by an unsupervised machine learning algorithm and use time-series analysis to flag anomalies in data.‘
Predicting future trends and tailoring content to fulfill readers’ and subscribers’ needs in advance is a powerful feature.
More from The Fix: How publishers are using data to fuel subscription growth
NYT Open sets the standard on informing on design, digital product building and user experience. The team has shown how they redesigned their popular newsletter The Morning, designed a new planning tool for their CMS or the story behind their homepage redesign.
Recently, The NYT Open team dived into a favorite topic of mine: how to recommend the right articles for readers to check out next. I’ve been in news a few years now and every year we see new startups pitch recommendation tools. I can’t remember one that worked well.
Assigning recommended articles by keywords and tags used by authors and editors has its issues. The blog says as much: interests might not correspond to tags for articles; tags represent a literal topic, while interests often represent a nuanced interpretation of that topic based on context.
The blog explains that building a machine-learning model that assigns interest labels to articles is a more powerful approach.
Surprisingly, the ending was kind of refreshing in the modern age of humanless automation:
‘We came to realize that even though our model outperforms the existing query-based system in many ways, it would be irresponsible to let it curate interests without human oversight.
Readers trust The Times to curate content that is relevant to them, and we take this trust seriously. This algorithm, like many other AI-based decision-making systems, should not make the final call without human oversight.
This interest classifier is already in use as one of a number of inputs our algorithms use to calculate article recommendations.‘
More from The Fix: The NYT subscriber strategy and why the model is hard to replicate
Hi! I'm David Tvrdon, a tech & media journalist and podcaster with a marketing background (and degree). Every week I send out the FWIW by David Tvrdon newsletter on tech, media, audio and journalism.