The Future of Sports News: A New Dataset and Approach

Imagine trying to turn live sports commentaries into news updates. It's a big task, and the data used now is often full of mistakes because computers collect and clean it. Also, there's a gap in knowledge between what's said live and what makes it into the news, which makes the job harder. In a new study, researchers introduced K-SportsSum. This dataset is special because it has a lot of data from many games – 7, 854 pairs of commentaries and news. To make it better, they cleaned the data by hand.

But what's really unique is that K-SportsSum also includes a big collection of information about 523 sports teams and 14, 724 players. This helps to fill in those knowledge gaps. The researchers also made a new tool that uses both live commentaries and this extra info to create sports news. They tested it on K-SportsSum and another dataset called SportsSum. Their model did really well and made news that people found more helpful.

Actions