Levels of Finite Clause Parsing: Implications and Applications

The aim of this paper is three-fold: (1) To point out the four types of inter-clause dependencies, thus levels of finite-clause parsing; (2) To discuss the essential role of the finite (and non-finite) clause structure in text processing, discourse parsing, text-to-speech modeling and prosody prediction; (3) To apply the Segmentation-Cohesion-Dependency (SCD) method based on marker classes, their semantic hierarchies, and marker sequences, combined with a machine learning technique, in order to obtain improved results on clause parsing of the Romanian Acquis corpus. A special emphasis of our analysis is devoted to the intonational role that the inter-clause lexical markers and their sequences are playing on the syntax-prosody interface (of Romanian), more precisely, for establishing the discursive-intonational contrastive foci and patterns that are necessary in prosody prediction.