- Up - | Next >> |

Embedding semantics: DRSs are viewed as partial models.

The idea behind the ``embedding semantics'' is that DRSs are viewed as partial models. Now, what does this mean? Basically, it means that a DRS is true with respect to a model (for example one which describes the real world) if it can be *embedded* in that model. An embedding succeeds if the discourse referents of the DRS can be mapped onto entities of the model's domain in such a way that all the conditions of the DRSs are fulfilled in this model.

Recall from Chapter 1 that we define models as ordered pairs , with a non-empty finite domain of entities, and an interpretation function mapping constants of the DRS language to elements or tuples of elements of . As usual, we have for an -place predicate symbol . Now consider the DRS for:

``A woman walks a dog.''

This DRS is true with respect to a model if we can find an assignment such that , and . This is the key intuition behind the partial match, and this will give us some idea as to how to define the model embedding semantics for DRSs in a formal way. Viewing assignments as partial functions from discourse referents to elements of , we come to the following definition:

- Up - | Next >> |

Aljoscha Burchardt, Stephan Walter, Alexander Koller, Michael Kohlhase, Patrick Blackburn and Johan Bos

Version 1.2.5 (20030212)