Glue Rules for Robust Chart Realization

This paper shows how glue rules can be used to increase the robustness of statistical chart realization in a manner inspired by dependency realization. Unlike the use of glue rules in MT—but like previous work with XLE on improving robustness with hand-crafted grammars—they are invoked here as a fallback option when no grammatically complete realization can be found. The method works with Combinatory Categorial Grammar (CCG) and has been implemented in OpenCCG. As the techniques are not overly tied to CCG, they are expected to be applicable to other grammar-based chart realizers where robustness is a common problem. Unlike an earlier robustness technique of greedily assembling fragments, glue rules enable nbest outputs and are compatible with disjunctive inputs. Experimental results indicate that glue rules yield improved realizations in comparison to greedy fragment assembly, though a sizeable gap remains between the quality of grammatically complete realizations and fragmentary ones.