Google Research

Underspecified Universal Dependency Structures as Inputs for Multilingual Surface Realisation

International Conference on Natural Language Generation (2018) (to appear)


This paper presents the datasets used in the First Multilingual Surface Realisation Shared Task (SR’18), describes in detail how they were created, and evaluates their quality. In addition, we examine (a) the NLG subtask of surface realisation itself, (b) the motivation for, and likely useful- ness of, deriving NLG inputs from annotations in resources originally developed for Natural Language Understanding (NLU), (c) whether the resulting inputs supply enough information of the right kind for the final stage in the NLG process, and more tentatively, (d) what role surface realisation is likely to play in the future in the NLG context.

Learn more about how we do research

We maintain a portfolio of research projects, providing individuals and teams the freedom to emphasize specific types of work