Parsing of partially bracketed structures for parse selection

Abstract

We consider the problem of parsing a sentence that is partially annotated with information about where phrases start and end. The application domain is interactive parse selection with probabilistic grammars. It is explained that the main obstacle is spurious ambiguity. The proposed solution is first described in terms of appropriately constrained synchronous grammars, and then in terms of a computational model for parsing. Experiments show the feasibility for a practical grammar.