The evolution of our local news geographic equity mapping & analysis tool

From simple points on a map, to adding filters, to incorporating census data about communities alongside community coverage.

In the fall of 2019, folks from the Brown Institute and Lenfest Local Lab came together to explore our shared questions about the relationship between geography, quality and experience in local news. After an initial review of research in the area of mapping and local news, we agreed that developing an automated approach to identifying and mapping locations found in news articles would deepen our research and product experimentation practices.

We believed that a tool that could plot locations in local news stories on a map would make a few significant contributions to our industry. First, a news mapping tool could help speed up and lower the costs of content auditing, especially audit work that examines the geographic representation of coverage as a whole. Also the project would provide a foundational model for the technology needs, strategy constraints and potential risks associated with personalizing news products for residents and communities that have the potential to cause harm.

Over time we built three different iterations of the tool, and each new version reflected an evolution in our thinking, or applied insights from collaborative conversations with stakeholders about the tool’s usefulness to a news organization. Here we’re sharing the logic behind each iteration, as well as our hopes for how The Inquirer and potentially other local news organizations could use the tool or build their own in the future.

Version 1: Initial Map Display

The first version of the mapping tool was the simplest interface, however building it from scratch required some of our most complex and cross-disciplinary thinking from our team. We developed a way to reliably access relevant stories from The Inquirer and then format them in a way that was friendly for extracting locations from the text. The team also defined an approach to geocoding the locations, so that latitude and longitude values for each location could be entered into mapping software and plotted on a map. To read more details about the initial technical approach, you can read more here: “Mapping Local News Coverage: Precise location extraction in textual news content using fine-tuned BERT based language.”

The initial map showed each location identified in a story (most stories mention multiple locations) as a dot on a map of the Philadelphia area. On hover, each dot revealed a pop-up window with information about the article. The pop-up included the article’s title and a link, the name of the location that was identified and the name of the location that was returned by the model, as well as the publish date and topic of the article. In addition, the map also featured color shading that reflected where the density of locations mentioned was high or where mentions were sparse, showing at a glance where there were clusters or gaps in coverage in the region.

Version 2: Adding filters

After sharing this version of the map with a few trusted colleagues, we started to hear similar questions. Could I see a map of just my desk’s stories? Could we see what a single author’s stories look like on a map? Could I change the date range to see stories just from a certain week or month? We knew that these types of questions could be easily answered by the map if we added simple filters, and so we added a handful of them in a panel to the left of the map— allowing for filtering by topic, author and date range.

In this version of the map, we also allowed users to change the way that geographic boundaries were drawn on the map of the region. A user could choose to look at coverage on a map on three levels of granularity: with large county-wide boundaries, with smaller neighborhood boundaries (in Philadelphia only) and with vey granular census tract boundaries.

Here we want to note that if a tool with these features were made available newsroom-wide, across a news organization or directly to the public, its creators would first want to consider and mitigate any risks. Even though each of the stories included on a map like this have been published, viewing them through the lens of a desk’s coverage or an individual author’s coverage without enough context and discussion could lead to could lead to harmful interpretations.

Version 3: Incorporating census data about communities

In the final version of our mapping tool we made major interface changes to achieve one of our goals for this third phase of the project, which was to explore ways to operationalize DEI practices in local newsrooms. We were able to explore this third phase thanks to support from Google’s GNI Innovation Challenge which in 2020, focused on aiding projects with DEI-related outcomes.

To help improve the map’s ability to help assess not just presence of coverage, but representation in coverage, we defined an approach for incorporating census data into the tool. We used the open-source, easy-to-use Census Reporter tool to determine which community data to include. (Census Reporter is an independent project aimed at making data from the American Community Survey more accessible.) The addition of data about residents was intended to enabled a sharpened focus on individuals within a community in the analysis made possible by the map. By focusing on the number of stories written per person in an area, rather than just the number of stories per area regardless of population density, you can get closer to thinking about coverage through the lens of person-by-person. To do this effectively and in a way that would scale in the time we had left, we narrowed down the scope of the map view to improve the speed and usability of the tool. This final version shows stories within one zip code at a time.

The other change we made in parallel was to include direct community collaboration in coversations about whether or not coverage of an area is inclusive and equitable. Soon we will publish insights from these efforts, which we undertook with Sabrina Vourvoulias, editor of the emerging Community & Engagement desk at The Philadelphia Inquirer, and folks from the New Kensington Community Development Corporation (NKCDC), a nonprofit the provides resources to residents and small businesses in the Kensington, Fishtown, and Port Richmond neighborhoods in Philadelphia.

Conclusion

Our hope is that the mapping tool and underlying technology we’ve built can be used in a number of ways going forward, beyond the life of our lab and this Google GNI-supported project.

Support ongoing work at The Inquirer. We hope that the tool can provide data and insights that support ongoing internal conversations and workshops about equity and representation within The Philadelphia Inquirer. We also believe that the tool can complement the important work being done by The Inquirer’s Community & Engagement in collaboration with a handful of folks who have recently joined the organization in community-focused roles. In addition, we believe that the tool provides an avenue for research and self-reflection that can advance the Inquirer’s own journalistic examination of the roots of systemic racism in Philadelphia, via the More Perfect Union project. And finally, we hope that the Inquirer will continue to update the mapping tool with new stories as they are published, allowing for ongoing geographic equity analysis of community coverage.

Support shared power with community through shared data. We also hope that a tool like this can be shared with community groups, policymakers and residents in Philadelphia, and in the future all places where local news shapes narratives about communities and impacts living conditions. By sharing access to this type of tool, newsrooms have the opportunity to share power and build or rebuild trust with communities via transparency, and enter into community accountability conversations that are tangibly more equitable and collaborative. We’ve found that in our community engagement work with the NKCDC, geography has been a tangible entry point for these challenging conversations.

Make location extraction and analysis more approachable and scalable. Even though our project is wrapping up, it feels like this work is just starting to take root. As advancements in natural language processing, deep learning, and geolocation techniques make this work more achievable, we hope to see more cross-disciplinary groups of technologists, designers, journalists, product managers, analysts and community groups teaming up to build tools that contribute to a better, more equitable and more transparent future for local news.

The Lenfest Local Lab was a multidisciplinary product and user experience innovation team. The Lab was founded within The Lenfest Institute for Journalism in 2018 before joining The Inquirer in 2021. The Lab ended their work in 2022.