Johannes Sommer, CEO of Berlin based company Retresco and computational linguist Mr. Prof. Dr. Hands Uszkoreit from German Research Center for Artificial Intelligence (DFKI), were interviewed by radio editor Manfred Kloiber from the Deutschlandfunk-program “Computer and Communication” about natural language generation, the current state of it as well as its perspectives.
Automated textgeneration based on data
In sports, economy or weather more and more articles are being published that are not written by an actual journalist but rather by so-called “software robots”. These automated generated articles are based on structured data, which is enriched with domain-specific knowledge as well as associated vocabulary. This way Retrescos software, called rtr textengine, transforms data into natural language texts within milliseconds.
While experts from Deutschlandfunk see this topic rather critically, different studies independently from another confirm, that readers consistently rate these texts positively. They also find it hard to differentiate them from editorial texts. This is due to the fact that technology in the field of artificial intelligence has made huge progress and will rapidly continue to develop within the next years.
Sebastian Golly, Head of Textgeneration at Retresco, explains the functionality of Retrescos rtr textengine by using football game reports as a reference. “I just took some match days of an Hamburg based amateur league and I am just going to generate some high quality game reports. Now you can see how these reports are being generated within seconds.”
In the background the system analyses associated data within seconds and transforms it into finished texts. In this way, rtr textengine generates recent pre- and post-game reports to every game in every league at a touch of button.
The possibility of real-time as well as on-demand coverage enormously enhances regional relevance of all kinds of portals. The result is a higher range, increased user engagement and a clearly enhanced visibility.