Beyond the Journal: The science of communication

Your next press release will not be written by a robot

06 Apr 2019
Fears that automation will take our jobs are rife, but press officers need not feel threatened by new tools that translate technical research into lay summaries.

By Ruth Francis

I recently attended a seminar that looked at tools we can use to help promote research. One of the panelists, Sabine Louët of Science Pod, sparked a discussion I had afterwards with two other press officers about the value we bring compared to what can be automated. 

Louët showed examples of news content that was written by Natural Language Processing, a sub-discipline of artificial intelligence. For example, in 2014, the first news article reporting a small earthquake in Los Angeles, California was generated in this way, and others were created from crawling data on street crime or climatic events. For those of us who struggle with headline writing, A*STAR and Singapore Press Holdings in Singapore may be able to help: they are using AI to make headlines more engaging without resorting to pure clickbait.

At Science Pod, they have software that can augment the work of a press officer by automatically creating a lay summary of a research article. In this way, it demystifies the original research much like a press release. 

Although next-generation summaries should be good enough to publish with little human intervention, press officers need not worry. These summaries are simply straight translations, not in-depth analyses or stories that we and journalists write.

A good press release should do much more than simply translate a research article. We can and should engage the reader and persuade them to find out more. We do this by adding detail gleaned by talking to the researchers, finding out why they did the work and what it adds to the field and beyond. 

Our ability to improve prose to clarify research, to tell stories, use metaphor and simile, add context and memorable quotes, all differentiate our releases from these summaries. Louët concludes, and I agree, that Natural Language Processing offers an efficient tool for press officers and media to understand research quickly, but the resulting summaries are just a first step. They leave the door wide open for us to create more informative and engaging content. 

As is the case with much artificial intelligence, some aspects of our work could be alleviated, leaving us freer to be more creative with the time we gain. 

Ruth Francis is a communications expert with 20 years of experience working in academia and publishing, including Springer Nature, BioMed Central, Cancer Research UK and King's College London.