Formal Issues in Natural Language Generation

Natural Language Generation (NLG) can be studied both empirically and formally. In recent times empirical research has tended to dominate the NLG research community, especially with the emergence of corpus and statistical techniques. However, formal research, based on proofs, formal models, and formal linguistics, can also contribute to the NLG research agenda. In this paper, we discuss what types of formal research are relevant for NLG, and we introduce the four papers in the Special Issue on Formal Issues in Natural Language Generation of the journal Research on Language and Computation in the light of this discussion.