Major advances in technology have made it possible to automatically pull locations out of local news stories, creating opportunities to be able to meaningfully personalize news for a reader’s neighborhood, town or block.
From journalist to editor to business manager, understanding geographical areas of newsroom coverage is critical to personalization and audience segmentation strategies—and is necessary to assess the reach, inclusivity, and utility of news. The Lenfest Institute and Brown Institute have developed an open-source automated approach to identifying and mapping locations in news articles using a mix of natural language processing, deep learning, and geolocation techniques.
Our guiding question is simple: what can be learned from seeing article locations on a map and how can that improve the business outlook for a local news organization? This question led to others: How does coverage lay out across a region? If there are gaps, can they be filled? Can this tool identify automated local angles to national stories?
In an era of reporting on COVID-19, news organizations know that a reader’s neighborhood has become their world. Hyperlocal news about testing sites, food banks, ventilators and volunteer opportunities are essential to staying healthy and informed.
This automated approach comes with challenges. To start, location identification is computationally challenging. Also locations in local news stories often need additional context to be accurate, including the names of small local businesses, buildings and street names. Next data quality is challenging, depending on the news organization and its content management system. Last, it will take further development to deploy the tool in other newsrooms.
Suggested Speaker(s)
- Sarah Schmalbach
Director of the Lenfest Local Lab, The Lenfest Institute for Journalism - Michael Krisch
Deputy Director, The Brown Institute