Automatic generation of Formula 1 feports

In this work, a new language generation system is proposed. Language generation in the area of sports reports is an active interest of the NLG community. However, there has been no research into the automatic generation of Formula 1 race reports in this area. The goal of the system that will be created is to produce short Formula 1 race reports. The system will create the reports with the help of a deep learning system called GPT-2, which will be finetuned to produce short Formula 1 reports. This research shows that it is possible to generate reports with similar quality of fluency, grammaticality, and cohesiveness as human-written reports However, it also shows the two biggest pain points for the system, which are the factuality and the repetitiveness.