Joost Zwarts

Vector Space Semantics: A Model-Theoretic Analysis of Locative Prepositions

This paper introduces a compositional semantics of locativeprepositional phrases which is based on a vector space ontology.Model-theoretic properties of prepositions like monotonicity andconservativity are defined in this system in a straightforward way.These notions are shown to describe central inferences with spatialexpressions and to account for the grammaticality of prepositionmodification. Model-theoretic constraints on the set of possibleprepositions in natural language are specified, similar to the semanticuniversals of Generalized Quantifier Theory.

Time and space in conceptual and logical semantics: the notion of Path

Jackendoffs treatment of Time and Space in his conceptual semantics is investigated in some detail. It is argued that his localistic system, which seems to account quite satisfactorily for directional phrases, can be made suitable for incorporation into a standard model-theoretic framework. This requires that Jackendoffs formalism, in particular his feature system, be interpreted in terms of semantic objects that are characterized settheoretically. This provides for a general treatment of sentences like John walked to the store and Judith ate three sandwiches, which behave similarly with respect to aspectuality.

A hierarchy of locations: evidence from the encoding of direction in adpositions and cases

Abstract The encoding of direction (place, goal, source, route) in systems of adpositions and local cases is not uniformly distributed over different locations (at, in, under), but can be shown to follow a hierarchical pattern. This pattern is compared with similar hierarchies proposed in the literature about the acquisition and typology of spatial language. Differences in semantic complexity and pragmatic salience between locations might explain why such a hierarchy exists.

Event semantics and abstract categorial grammar

Common versions of event semantics do not naturally explain the obligatory narrow scope of existential quantification over events, or the typically event-oriented modification by adverbials. We argue that these linguistic properties reflect a distinction between overt arguments and purely semantic slots like the event argument. The distinction is naturally captured in Abstract Categorial Grammar (ACG) [1, 2, 3, 4], which manipulates pairs of forms and meanings, a.k.a. linguistic signs. The signs pheno-type defines syntactic arguments and the signs semantic type standardly defines semantic arguments. Both these concrete types are standardly derived by induction on the structure of one abstract type (category) of the sign, by assigning pheno-level and semantic types to basic abstract types. We assume that semantic event arguments are only introduced by the (basic) result type of the verbs abstract type, whose pheno-level type is standardly a string. Consequently semantic event arguments lack a correlate in the verbs pheno-type. Both narrow-scope existential quantification over events and the orientation of event modifiers follow rigorously from this assumption. Based on this architecture, we develop simple accounts of adverbial modification, nominalization and passive constructions in an ACG fragment.

The distribution of labor between adpositions and case within complex spatial PP's

Abstract This paper discusses a cross-linguistic sample of spatial PPs in languages that both have adpositions and case. It is shown that the distribution of labor within these potentially complex PPs follows from two general principles only. According to the principle of Grammaticalization, less frequent meaning elements should never be expressed by more grammatical means than more frequent ones. According to the principle of Compositionality, the syntactic construction should reflect the order of semantic function application. The only viable spatial PP constructions according to these principles are those constructions in which the P simultaneously expresses configuration and directionality, and constructions in which the P expresses configuration and the case marker on the P directionality.

Semantic markedness in gender opposition, blocking and fossilization

Abstract The semantics of natural gender in animate nouns is modeled in the framework of bidirectional Optimality Theory (OT). This allows for the interaction of lexical, conceptual and contextual constraints and for a straightforward treatment of the effect of blocking in this domain. Two versions of bidirectional OT are discussed and related to each other in terms of Blutner’s (2007) notion of fossilization.

Spatial semantics: Modeling the meaning of prepositions

Reference to space, especially with prepositions, plays a central role in natural language and is receiving more and more attention over the past decades. One line of research uses formal semantic modeling, using topological and other geometrical concepts such as regions, vectors, and paths. Another line of research has drawn attention to the role of function, force-dynamics, polysemy, prototypes, and crosslinguistic variation in this domain. This paper gives an overview of both lines of research and argues that a synthesis is possible, based on a proper division of labor between semantics and pragmatics, richer ontologies, and a perspective on categorization that uses conceptual spaces and semantic maps.

Implicit arguments: event modification or option type categories?

We propose a unified syntactic-semantic account of passive sentences and sentences with an unspecified object (John read). For both constructions, we employ option types for introducing implicit arguments into the syntactic-semantic categorial mechanism. We show the advantages of this approach over previous proposals in the domains of scope and unaccusatives. Unlike pure syntactic treatments, option types immediately derive the obligatory narrow scope of existential quantification over an implicit arguments slot. Unlike purely semantic, event-based treatments, our proposal naturally accounts for syntactic contrasts between passives and unaccusatives, as in the door *(was) opened by John.

Shifting animacy shifts

De Swart and de Hoop (henceforth dS&dH) make the suggestion that the linguistic distinction between animate and inanimate categories should be brought within the scope of the formal semantic theory of types and type shifts, overt and covert. I would like to work out this interesting idea a bit more. The result might be different from what dS&dH had in mind, although it is hopefully in the same promising spirit. Let me start with their idea that the type e, corresponding to the domain of entities E, has two subtypes: a (for dogs and men, for instance) and i (for breads and cups, for instance). The type distinction between a and i seems straightforward enough, but it does not provide a good foundation for type shifts in the formal sense. Such type shifts require some sort of structure, which the domain of entities all by itself does not have. Once we have complex types, like 〈e,t〉 (domain of sets) and 〈〈e,t〉,t〉 (domain of quantifiers), we also have useful shifts, like ident, which can shift any entity x to the singleton set {x} (Partee 1987). It is also possible to create more structure within the domain of entities, for example, by having kinds that are realized as individual objects (Carlson 1977) or individual objects that consist of stuff (Link 1983). Once we have such a richly structured domain of entities, it becomes possible to define shift between properties and kinds and mass and count (e.g., Chierchia 1998). It is not clear how we could structure the domain E in such a way that animate and inanimate entities are systematically related to each other. It is true that we might imagine some sort of function, call it STATUE, from animates to their inanimate representation (i.e. of type 〈a,i〉) as well as a function of type 〈i,a〉, say WAND, that takes an inanimate (e.g. a cup) and makes it a living thing (in a fairy tale context). Such functions could at first sight accomplish the required covert conceptual shifts in examples (3) and (4) from dS&dH (with my analytical embellishments):

Conceptual Spaces, Features, and Word Meanings: The Case of Dutch Shirts

This paper explores how a conceptual space for the representation of word meanings can be constructed and visualized for one particular domain, namely Dutch words for different types of shirts. It draws on earlier empirical corpus-based research that has identified different features for uniquely describing each of these types and different ways in which they are lexically described in fashion magazines. The present study defines a metric that makes it possible to construct a feature-based space in which the extension of each of the Dutch shirt terms can be visualized and in which it is possible to study the distribution of words and the validity of different constraints on that distribution: conjunctivity, convexity, connectivity, coherence, and centrality. Although the paper concludes that definite conclusions about these constraints are only possible on the basis of more complete lexical datasets, it demonstrates the potential of the conceptual space approach for studying word meanings.