Murrinh-Patha Complex Verbs : Syntactic Theory and Computational Implementation

Abstract

Murrinh-Patha is a non-Pama-Nyungan language spoken in and around Wadeye in the Northern Territory of Australia. It is a highly polysynthetic language with a complex verbal structure. The Murrinh-Patha verb morphology is not only complex due to the large number of morphemes that a verb may consist of, but also due to a high degree of syncretism that these morphemes display. Additionally, complex interdependencies exist between the morphemes. The system is further complicated by a complex interface to the syntax and semantics, as no one to one correspondence between the morphological encoding and its syntactic and semantic interpretation exists. Murrinh-Patha morphology and its interface to syntax and semantics thus poses many challenges both for theoretical analysis and computational implementation.The main focus of this work is on one special interdependency in detail, namely on the interplay of two morphemes which together contribute the core semantic meaning of the verb. These morphemes are referred to as classifier stem and lexical stem and their combinations are considered morphological complex predicates.Complex predicates have been discussed for many different languages in a range of diverse frameworks. In most accounts of complex predication, the constraints on which elements may combine to form a valid complex predicate are cast within argument structure constraints. This thesis argues that in Murrinh-Patha, semantic concepts which would normally be considered encyclopedic knowledge play a prominent role in the selection process of the combinatory possibilities, irrespective of the number of arguments involved. To model these combinations, this thesis proposes a new way of analyzing complex predicates by combining Jackendoff’s (1990) Lexical Conceptual Structures (LCSs) with a relational approach to lexical semantics. The semantic concepts that restrict combinatory possibilities are modeled with hierarchies of selectional restrictions. These hierarchies are then used to derive the argument structure of the complex predicates in the form of LCS blueprints. Only the combination of both, I claim, can explain both the combinatory restrictions and the mapping to syntax found in Murrinh-Patha.The second part of the thesis is concerned with the computational implementation of Murrinh-Patha, with a special focus on the implementation of Murrinh-Patha verbs. The implementation of the Murrinh-Patha morphology is carried out with the Xerox finite-state technology tools XFST and LEXC (Beesley & Karttunen 2003) and is then used in a corpus study of a small Murrinh-Patha bible corpus with the aim of extracting new verbal candidates. The morphology is also used in a computational implementation of some parts of the Murrinh-Patha syntax, using the grammar development platform XLE (Crouch et al. 2015). The XLE grammar is then used in a range of applications for beginning learners of Murrinh-Patha, such as an electronic dictionary and a small English to Murrinh-Patha translation system.