Project B4 (2006-2018)

Principal Investigator: Antje Roßdeutscher and Hans Kamp
Researchers: Tillmann Pross and Boris Haselbach
Associated Ph.D. Student: Sylvia Springorum

B4 investigates the composition of meaning from sublexical units all the way up to that of sentences and discourse. We investigate the source and resolution of ambiguity in an architecture of syntactically driven composition that combines a syntax based on the principles of Distributed Morphology with semantic representation formalisms from Discourse Representation Theory. In phase 3 additional attention will be paid to ontological building blocks of meaning and to parallels at the syntax-semantics interface in the analysis of argument structure, Aktionsart and case assignment across the verbal, nominal and prepositional domain. Our theories about -ung-nominals and particle and prefix verbs will be empirically validated in the silver standard program of the SFB.

In phase 3 B4 will shift its focus to general questions about the nature and architecture of a syntax-semantics interface which reaches from sublexical structure all the way to that of sentence and discourse. As during phases 1 and 2, we will combine a syntax based on the principles of Distributed Morphology with a semantics which uses representations from a formalism based on Discourse Representation Theory.
These commitments give rise to a variety of questions. The most important of these for the work planned for phase 3 are: What is the relationship between semantics and ontology? More specifically: can we model the composition of word meaning as the syntactically driven organization of ontological units, which are directly assigned to roots or which result through application of compositional operations to other such units? In other words, how does syntax guide the construction of ontological complexes out of simpler ontological building blocks in the verbal, nominal, prepositional and adjectival domain? And what, in particular, can we learn from the answers to these questions about the ontological configurations expressed by prefix and particle verbs and by -ung nominalizations?
Under-determinacy and ambiguity are pervasive in sublexical semantics, as is the efficiency of the specification and disambiguation mechanisms by which most sublexical ambiguities and under-determinacies are resolved (usually at an 'early', very 'local' level). We have made some headway with the task of identifying these 'incremental specification' mechanisms. But what we have done will need refinement and adjustment in the light of results from our explorations of the relations between sublexical semantics and ontology.
A third aim will be empirical validation against large data of the hypotheses to which our work has led us so far and those that we will arrive at in the course of phase 3. To this end we will collaborate with the projects B9, D12 and INF, which we will rely on for the necessary computational tools and infrastructure. New will be an emphasis on non-canonical data. For instance, we will investigate 'non-canonical', i.e. dispositional -ung nominalizations and their base verbs. And we will relate our work on -ung nominalizations to that proposed in B9, exchanging the results of our logic-based approach to semantics with those of B9's Distributional Semantics approach. Also, we will, in collaboration with D12, test the productivity of particle and prefix verb formation by eliciting judgments from theoretically unbiased speakers about the possible meanings of made-up verbs.