Hamburg Working Papers on Language Production - II

What's up with the printer? Context relative presentation of conceptual structure and its prosodic realization in a language production system.. The papers presented in this issue show some results of current work being performed in the SYNPHONICS * project group. They continue what gradually becomes a tradition: to regularly report in a compact form about the actual research that centers around different aspects of language production. The SYNPHONICS project adopts a cognitive approach to a computational linguistic model of language production that combines results from psycholinguistic research about the time course of the human language production process with recent developments concerning the representation of semantic, syntactic and phonological knowledge in theoretical linguistics. The aim of the project consists of the development of a system that covers the incremental generation of utterances from pre-linguistic conceptualization to the formation of phonological structures, which are in turn interpreted phonetically, yielding an articulatory input to a speech synthesis module. A central claim of SYNPHONICS is that procedural aspects of language production are distinguished from declarative representations of linguistic knowledge. The declarative linguistic knowledge base encodes semantic, syntactic, and phonological (i.e. segmental and prosodic) information in a special variant of a Head-driven Phrase Structure Grammar, in this case, for German. Articulatory gestures are accordingly represented in a feature-based notation. The overall procedural design of the system is reflected by an architecture that organizes incremental processing of partial information. Conceptual information that enters the formulator is processed in parallel on a lexical and a structural track, reunified by a subsequent processing unit. Within this global setting the phonological planning component performs lexical phonological processes as well as abstract prosodic planning processes. Motivated by results from empirical research in psycholinguistics (for example, in the field of speech error investigation), the output of the lexical phonological component is separated into a segmental and a prosodic part. Only the latter serves as input to the abstract prosodic planning component. Among the linguistic phenomena that are analyzed within the SYNPHONICS framework, emphasis is placed on investigations concerning the information structure of an utterance in terms of focus-background and topic-comment. We argue in particular that meaning distinctions triggered by changes in information structure can be reflected by prosodic means without any support from syntax. Therefore, SYNPHONICS conjectures a direct semantics/phonology interface in addition to the commonly assumed syntax/semantics and syntax/phonology interfaces. This enables the phonological component to directly access semantic information …