|Title:||Lexical Semantics and Linking in the Hierarchical Lexicon||Add Dissertation|
|Author:||Anthony Davis||Update Dissertation|
|Email:||click here to access email|
|Institution:||Stanford University, Department of Linguistics|
|Linguistic Subfield(s):||Semantics; Syntax;|
|Abstract:||This thesis sets out a model of the interface between semantic roles and syntactic arguments, within the framework of Head-Driven Phrase Structure Grammar (HPSG). This model is constraint-based, semantically grounded, and constructed using multiple inheritance of feature structures. The use of feature structures keeps the model conceptually simple, and the constraint-based approach is in keeping with the HPSG outlook on grammar. In presenting this model, I also demonstrate that no special mechanisms or components of grammar are required to model the mapping between semantic roles and syntactic arguments. The independently motivated mechanisms of constraint-based grammatical frameworks, such as HPSG, serve the purpose.
The model is semantically grounded in the sense that the semantic roles of a predicator are classified by the semantic entailments that hold of the participants in the event or state the predicator denotes. In this respect it differs from theories that employ thematic roles, tacitly assigning a unique thematic role to each of a predicators arguments but leaving the semantic conditions for such assignment unclear. For the purposes of describing how semantic roles are realized as syntactic arguments, I advocate associating classes of semantic entailments with proto-roles. Each proto-role corresponds to sets of entailments that act alike in the mapping from semantic roles to syntactic arguments. Proto-roles are represented as attributes within the lexical semantics of predicators.
The model is constraint-based because it treats the patterns we observe in the mapping between semantic roles and syntactic arguments as the result of interacting linking constraints on lexical entries in natural languages. Linking constraints can be kept simple because each one is a partial specification of the mapping between lexical semantics and subcategorization. Several constraints can apply to a lexical entry, which must satisfy them all, and various unattested patterns are ruled out. The structure of the lexical hierarchy ensures that the constraints applicable to a given predicator will indeed apply. I apply the model to a wide range of verbs, including passive verbs, which appear on the surface to violate linking constraints, and to argument structure alternations.