Generating referring expressions containing Quantifiers
Sebastian Varges, Kees van Deemter · 2005
the generating capability of algorithms in this area. This paper asks whether the models underlying these proposals can still be used if even more complex referring expressions are generated. To discuss this issue, we will investigate a variety of referring expressions that pose difficulties to current generation algorithms. In particular, we will discuss the difficulties associated with quantified referring expressions (such as ‘those women who have fewer than two children’, or ‘the people who work for exactly 2 employers’) and explain how they can be generated by extending the inference-based approach described in (Varges, 2004). 1 GRE: descriptions using conjunctions of atoms, and beyond The Generation of Referring Expressions (GRE) is a key task in most Natural Language Generation (NLG) systems, which is usually viewed as part of microplanning (Reiter and Dale 2000, Chapter 5). Simply put, GRE amounts to finding suitable descriptions in natural language that allow a