Supertagging with combinatory categorial grammar for dependency parsing

Download
2014
Akkuş, Burak Kerim
Combinatory Categorial Grammar (CCG) categories contain syntactic and semantic information. CCG derivation trees can be used in extracting partial dependency structures by providing the missing information in order to build complete dependency structures. Therefore, CCG categories are sometimes referred to as supertags. The amount of information encoded in supertags makes it possible to create very accurate and fast parsers as supertagging is considered ``almost parsing''. In this thesis, a maximum entropy based part of speech tagger is presented to improve the performance of CCG supertagging and another maximum entropy classifier is implemented with additional features for supertagging. Morphological features of words of an agglutinative language such as Turkish are used in order to improve the accuracy of POS tagging and supertagging processes. This indicates direct relationships between morphemes and lexical categories. The effects of using the improved supertagger are tested on dependency parsers by means of using supertags as rich parts of speech tags. Additionally, using POS taggers that assign multiple part of speech tags to the ambiguous words is suggested as another potential improvement for supertaggers.

Suggestions

A type system for combinatory categorial grammar
Erkan, Güneş; Bozşahin, Hüseyin Cem; Department of Computer Engineering (2003)
This thesis investigates the internal structure and the computational representation of the lexical entries in Combinatory Categorial Grammar (CCG). A restricted form of typed feature structures is proposed for representing CCG categories. This proposal is combined with a constraint-based modality system for basic categories of CCG. We present some linguistic evidence to explain why both a uni cation-based feature system and a constraint-based modality system are needed for a lexicalist framework. An implem...
Implementing the type-raising algorithm by grammar compiling
Demir, Oğuzhan; Bozşahin, Hüseyin Cem; Department of Cognitive Sciences (2019)
Type-raising is part of theory of Combinatory Categorial Grammar, by which all arguments including complements are type-raised. Generating type-raising rules in an automatic manner in the compile-time via a simple tool would make experimenting with Combinatory Categorial Grammar faster, allowing control on each run. In this study, created tool is tested with various grammars including large scale Eve database, giving results in O(N) where N is the number of verbs in the grammar.
CROSS-LEVEL TYPING THE LOGICAL FORM FOR OPEN-DOMAIN SEMANTIC PARSING
Öztürel, İsmet Adnan; Bozşahin, Cem; Department of Cognitive Sciences (2022-8-29)
This thesis presents a novel approach to assigning types to expressive Discourse Representation Structure (DRS) meaning representations. In terms of linguistic analysis, our typing methodology couples together the representation of phenomena at the same level of analysis that was traditionally considered to belong to distinctive layers. In the thesis, we claim that the realisation of sub-lexical, lexical, sentence and discourse-level phenomena (such as tense, word sense, named entity class, thematic role, a...
Gradient characteristics of the unaccusative/unergative distinction in Turkish : an experimental investigation
Acartürk, Cengiz; Zeyrek Bozşahin, Deniz; Department of Cognitive Sciences (2005)
This thesis investigates the gradient behaviour of monadic intransitive verb classes in Turkish, under an aspectual classification of the unaccusative/unergative verb types, namely The Split Intransitivity Hierarchy. This Hierarchy claims that intransitive verb types are subject to gradient acceptability in certain syntactic constructions. The methods used in judgment elicitation studies in psychophysics, such as the magnitude estimation technique have recently been adapted to be used in capturing gradient ...
Measuring empirical bias toward ergativity and accusativity
Şakiroğulları, Çağrı.; Bozşahin, Hüseyin Cem; Department of Cognitive Sciences (2019)
Starting with six word order possibilities for a basic transitive clause, exposed data can bias English children to the point of making a categorial distinction for SVO. However, for an SVO language there are two categorial possibil- ities, an ergative and an accusative one. The acquisition of ergativity and accusativity is a complex phenomenon, since they both have similar phono- logical structure for the baby. We show, using Eve database of CHILDES, that these possibilities can be acquired from exposed da...
Citation Formats
B. K. Akkuş, “Supertagging with combinatory categorial grammar for dependency parsing,” M.S. - Master of Science, Middle East Technical University, 2014.