Monok featured in University of Brighton research on automated journalism
Anya Belz examines Monok’s automated journalism and compares a published article with human-written news using the GLTR detection tool.
Monok is featured in University of Brighton researcher Anya Belz’s 2019 paper, Fully Automatic Journalism: We Need to Talk About Nonfake News Generation, alongside the UK news service RADAR.
For a company founded in Stockholm in 2016, it is a meaningful point in a longer journey. Monok grew out of research into language processing, machine learning and human-computer interaction. Our focus has been turning those ideas into systems that can carry information through a working publishing process.
In production, not just in a demo
Belz places Monok among the early news-generation systems already operating commercially. She describes its combination of extraction, summarization and rewriting, and highlights its ability to produce factual content economically while meeting publishers’ search-visibility needs.
How Monok compared with human-written news
The paper includes a GLTR comparison between a published Monok article and a similar human-written Fox News article. In that example, Belz reports that Monok’s output appeared slightly more likely to be human-authored than the human-written reference.
This is the specific comparison presented in Figure 3: an examination of the language patterns in those two articles using a text-generation detection tool.
More room for original journalism
Journalism.co.uk covered the paper on November 11. In that interview, Belz discussed the value of automation for local and regional publishers: reducing the cost of meeting content demand can help newsrooms survive and give their staff more space for local investigative reporting.
That practical publishing need is central to our work. Finding relevant information, connecting reports about the same subject and preparing material for publication all take time. Our early systems bring those steps together, rather than treating the generation of text as a separate task. The aim is a process that a publishing team can use repeatedly as part of its everyday work.
The University of Brighton lists the paper as published on October 25, 2019, following its presentation at the Conference for Truth and Trust Online in London.