Universität zu Köln

Bogner, Florian (2020). The Morphosyntactic Parser: Developing and testing a sentence processor that uses underspecified morphosyntactic features. PhD thesis, Universität zu Köln.

Preview

PDF Bogner (2020) – The Morphosyntactic Parser.pdf Download (2MB) | Preview

Abstract

This dissertation presents a fundamentally new approach to describe not only the architecture of the language system but also the processes behind its capability to predict, analyze and integrate linguistic input into its representation in a parsimonious way. By the example of morphosyntax, underspecified case, the use of decomposed, binary case, number and gender features to account for syncretism, will offer insights into both: Carrying over this idea to language processing raises the question whether the language system—limited in its storage capacity—makes use of similar means of representational parsimony during the processing of linguistic input. This thesis will propose a processing system that is tightly related to the aforementioned architectural assumptions of morphosyntactically underspecified lexical entries as a parsimonious way of representation. In that sense, prediction is viewed as the language system’s drive to avoid feature deviance from one incrementally available linguistic element to another subsequentially incoming one. In this way, the parser’s goal is to maintain minimal feature deviance or at best feature identity to keep processing load as low as possible. This approach allows for position-dependent hypothesis with regard to the expected processing load. To test the processor’s claims, the electrophysiological data of a series of event-related brain potential (ERP) experiments will be presented. The results suggest that with the input’s increased feature deviance the amplitude of an ERP component sensitive for prediction error increases. In comparison to that, elements that rather maintain feature identity and that do not lack or introduce additional features to the analysis do not increase processing difficulty. These results indicate that the language processing system uses the available features of morphosyntactically underspecified mental entries to build up larger constituents. The experiments showed, that this buildup process is determined by the language system’s drive to avoid feature deviance.

Item Type:	Thesis (PhD thesis)
Creators:	Creators Email ORCID ORCID Put Code Bogner, Florian bogner.f@googlemail.com UNSPECIFIED UNSPECIFIED
URN:	urn:nbn:de:hbz:38-113863
Date:	2020
Language:	English
Faculty:	Faculty of Arts and Humanities
Divisions:	Faculty of Arts and Humanities > Fächergruppe 3: Deutsche Sprache und Literatur > Institut für Deutsche Sprache und Literatur I
Subjects:	Language, Linguistics
Uncontrolled Keywords:	Keywords Language Morphosyntax English Morphosyntactic features English Morphosyntaktische Merkmale German Syncretism English Synkretismus German Underspecification English Unterspezifikation German Incremental sentence processing English Inkrementelle Satzverarbeitung German Incremental processing English Inkrementelle Verarbeitung German Features English Merkmale German Subject object ambiguities English Subjekt-Objekt-Ambiguitäten German Subject preference English Subjektpräferenz German ERP English EKP German Event related potential English Ereigniskorreliertes Potential German Event related potentials English Ereigniskorrelierte Potentiale German N400 UNSPECIFIED LAN UNSPECIFIED P600 UNSPECIFIED
Date of oral exam:	22 June 2020
Referee:	Name Academic Title Schumacher, Petra Prof. Dr. Gutzmann, Daniel PD Dr. Bunčić, Daniel Prof. Dr.
Refereed:	Yes
URI:	http://kups.ub.uni-koeln.de/id/eprint/11386