Deadly drug combos uncovered with machine learning

 

The Chicago’s Tribune’s look at deadly Rx drug combinations relied on machine learning.

Reporters teamed up with data scientists who had access to prescription drug data and identified commonly-prescribed meds that could sicken or kill patients. Note: The journalists could not perform the analysis themselves because only the researchers had access to the detailed prescription data.

Machine learning helps ID Supreme Court gatekeepers

 

Reuters used machine learning for its Echo Chamber investigation, which examined the outsized role that a small number of Washington, D.C., lawyers had in getting the U.S. Supreme Court to hear appeals.

Reuters analyzed the text of petitions to the court, which it could access through the affiliated Westlaw court-reporting service.

Artificial intelligence powers interactive storytelling

Here’s a great instance of how journalists can use artificial intelligence (AI) in video storytelling.

China Daily’s Asia Weekly produced this interactive video interview with Nepali trans activist Bhumika Shrestha using IBM Watson AI

Read more about the interview and coding process here.

Graph databases help probe offshore investments

For the past few years, journalists with the International Center for Investigative Journalists (ICIJ) have teamed up to report on global corruption.

Noteworthy investigations include the center’s Offshore Leaks investigations, based on leaked documents detailing investments in tax havens.

 

The center made big news with its Panama Papers stories, which won a 2017 Pulitzer Prize for Explanatory Reporting. 

ICIJ used Neo4j and Linkurious web visualization tool to help its network of more than 100 journalists worldwide explore the graph database holding information about connections between people, companies and hidden investments.

Machine learning helps detect abusive doctors

For its series about sexually-abusive doctors across the United States, the Atlanta Journal Constitution needed to build its own database. No one centralized source collected that information, so reporters scraped state government websites to harvest medical board disciplinary information.

Then reporters applied machine learning to analyze more than 100,000 cases and score each on the probability that sexual abuse had occurred.