Supporting journalism by combining neural language generation and knowledge graphs

Natural Language Generation is a field that is becoming relevant in several domains, including journalism. Natural Language Generation techniques can be of great help to journalists, allowing a substantial reduction in the time required to complete repetitive tasks. In this position paper, we enforce the idea that automated tools can reduce the effort required to journalist when writing articles; at the same time we introduce GazelLex (Gazette Lexicalization), a prototype that covers several steps of Natural Language Generation, in order to create soccer articles automatically, using data from Knowledge Graphs, leaving journalists the possibility of refining and editing articles with additional information. We shall present our first results and current limits of the approach, and we shall also describe some lessons learned that might be useful to readers that want to explore this field.