Nominalization and Argument Structure: An Experiment with the NOMLEX Database

Download

10497417.pdf

Date

2022-8-31

Author

Kızıldemir, Melis

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

139
views

36
downloads

This study presents a method for predicting syntactic structures of deverbal event nominals that are derived from verbs used in their transitive meaning. The study takes a data-driven approach and fine-tunes pre-trained deep learning models. The problem is treated as a 15 class multi-label classification task and NOMLEX is used to create the classes. In order to leverage the language models learned by pre-trained models, sentences available in corpora which include the verb that the deverbal nominal is derived from is used to fine-tune the models. As a result of this, selectional preferences of verbs are provided to the model implicitly. Sentences where the verb is used in its transitive meaning are filtered from the corpora using EasyCCG CCG parser. The DeBERTa model scores 66.4% in sample based accuracy and 77.9% in sample based F1score, and the RoBERTa model scores 75.0% in label based F1 score. When the models are evaluated on classes, at least one model performs better than baseline in 10 out of 15 classes. Deverbal nominal syntactic structures that only realize an argument in the possessive determiner position, two out of three syntactic structures that realize the subject in the posessive determiner position and two out of three syntactic structures that realize the subject in the noun modifier position are among unsuccessfully learned classes. It is concluded that determiners of an argument's realization in a syntactic position is dependent on both the argument and the syntactic position in question along with realization of other arguments. The presented method was able to learn these determiners to various extents.

Subject Keywords

nominalization, deep learning

URI

https://hdl.handle.net/11511/99517

Collections

Graduate School of Informatics, Thesis

Suggestions

OpenMETU
Core

Paracompositionality, MWEs and argument substitution Bozşahin, Hüseyin Cem (2018-08-11) Multi-word expressions, verb-particle constructions, idiomatically combining phrases, and phrasal idioms have something in common: not all of their elements contribute to the argument structure of the predicate implicated by the expression. Radically lexicalized theories of grammar that avoid string-, term-, logical form-, and tree-writing, and categorial grammars that avoid wrap operation, make predictions about the categories involved in verb-particles and phrasal idioms. They may require singleton types,...
Argumentation in undergraduate math courses A study on definition construction Ubuz, Behiye; Bülbül, Ali (null; 2013-07-01) The purpose of this study is to analyzethe complex argumentative structure in undergraduate mathematics classroom conversations during definition construction by taking into consideration students’ and teacher’ utterances in the classroom using field-independent Toulmin’s theory of argumentation . The analyses contributed to an emerging body of research on classroom conversations.
ARGUMENTATION IN UNDERGRADUATE MATH COURSES : A STUDY ON PROOF GENERATION Ubuz, Behiye; BÜLBÜL, ALİ (2012-07-22) The purpose of this study is to analyze the complex argumentative structure in undergraduate mathematics classroom conversations by taking into consideration students' and teacher' utterances in the classroom using field-independent Toulmin's theory of argumentation. The analyses contributed to an emerging body of research on classroom conversations.
Learning transferability of cognitive tasks by graph generation for brain decoding Coşkun, Bilgin; Yarman Vural, Fatoş Tunay; Department of Computer Engineering (2021-12-10) Brain decoding involves analyzing the cognitive states of human brain by using some statistical techniques in order to understand the relations among the cognitive states, based on neuroimaging data. A very powerful tool to acquire the brain data is functional magnetic resonance images (fMRI), which generates three-dimensional brain volume at each time instant, while a subject performs a cognitive task involving social activities, emotion processing, game playing, memory etc. However, it is very difficult a...
Probability learning in normal and parkinson subjects: the effect of reward, context, and uncertainty Erdeniz, Burak; Gökçay, Didem; Department of Cognitive Sciences (2007) In this thesis, the learning of probabilistic relationships between stimulus-action pairs is investigated under the probability learning paradigm. The effect of reward is investigated in the first three experiments. Additionally, the effect of context and uncertainty is investigated in the second and third experiments, respectively. The fourth experiment is the replication of the second experiment with a group of Parkinson patients where the effect of dopamine medication on probability learning is studied. ...

Citation Formats

M. Kızıldemir, “Nominalization and Argument Structure: An Experiment with the NOMLEX Database,” M.S. - Master of Science, Middle East Technical University, 2022.