A robot journalist called Dreamwriter, designed by Chinese technology and gaming giant Tencent, churned out a 916 word business copy titled ‘China’s August consumer price index’ in just 60 seconds last week, sending the country’s media into a tizzy. The article was written in Chinese, had zero errors and was released through the company’s instant messaging service. While this is not the first time that an automated writer has replaced a journalist in the newsroom — the US has been doing it consistently since 2012 — the development has the potential of changing the dynamics of news reporting.
HOW DOES IT WORK?
The first story writing algorithm called Tale-Spin was was developed in 1977 at Yale University. It used knowledge about “problem solving, physical space, inter-personal relationships, character traits, bodily needs and story structure” to write copies. Since then, a slew of algorithms has been developed to power robot journalists. University of Maryland Professor Nick Diakopoulos divides its functioning into five stages:
News beats such as weather, sports and business have a large amount of standardised data that needs to be first ingested or read by the algorithm of the robot journalist. The data needs to be in a “clean and comprehensive” format for the robot to store the information. Apart from numbers the robots can ingest quotes from analysts too.
Next the robot spots “newsworthy” points in the story, for which the algorithm has built-in “criteria”. From its trove of data, the algorithm begins isolating the points like “maximum, minimum, biggest, changes in value, crossing of a benchmark ” etc. An infinite number of criteria can be built into the statistics of the algorithm.
The robot then needs to give an “angle” to the story. The angles, which are essentially “patterns of events, circumstances, entities, and their features”, are picked up from a pre-authored library. An angle can be anything from “biggest market slump ever” to “best performance on the field”. The angles are ranked from 1-10, and the robot selects the top-ranking angle that “fits the story”.
The last stage is writing the news piece for which the algorithm has specific templates. The templates have the “who, what, where, when, why and how” of a story. A “phrasal generation” programme helps in sentence formation.
LINK TO STORY POINTS:
Then all data needs to be linked to specific story points, like names of people, locations etc. The robot can also pull out information about a businessman or sportsperson from the Internet — like which place he belongs to, how old he is etc.
OTHER ROBOT JOURNALISTS
Quill/Narrative Science: Used to write sports and finance stories used by Forbes magazine and Big Ten Network
Quakebot/ Los Angeles Times: Writes news about earthquakes and live tweets latest developments
Wordsmith/Automated Insights: Produced one billion news copies last year. Used by Associated Press to write 3,000 copies every quarter by adopting its news style guide.
1) Speed is the biggest advantage of a robot journalist, as it can produce thousands of copies in seconds.
2) It gives accurate information, as the algorithms are designed to “catch errors and learn from mistakes”.
3) It can easily handle complicated data processing tasks, facts and figures, leaving it to human reporters to to add the analysis.
4) They also come cheap – it takes as little as $7 to produce an automated copy in the US.
5) They are a boon for the employers, taking no holidays and never missing a deadline.
SHOULD JOURNALISTS BE SCARED?
In China, reporters working for government-run newspapers regularly put out press releases without any analysis or investigation. This form of journalism does face a threat from robots that produce similar copies in much lesser time and money. But for other parts of the world, a robot journalist can at best serve as an assistant. They are not equipped to write feature or human interest stories and lack analytical skills. They cannot read PDF files or Access spreadsheets.