Korean Null Object Resolution: An Unsupervised Machine Learning Approach

We aim to develop an unsupervised machine learning (ML) model for co-reference resolution of anaphoric phonologically-null objects (APNOs) in radical pro-drop East-Asian languages. Focusing on Korean, we adopt as input the APNOs in the discourse-theoretically annotated corpus. Specifically, the corp...

Full description

Saved in:
Bibliographic Details
Published inInternational Information Institute (Tokyo). Information Vol. 20; no. 6B; pp. 4559 - 4568
Main Author Kim, Euhee
Format Journal Article
LanguageEnglish
Published Koganei International Information Institute 01.06.2017
Subjects
Algorithms
Annotations
Artificial intelligence
Clustering
Computer science
Datasets
Engineering
Linguistics
Machine learning
Theory
Training
Unsupervised learning
Asia
Online AccessGet full text

Cover

More Information
Summary:We aim to develop an unsupervised machine learning (ML) model for co-reference resolution of anaphoric phonologically-null objects (APNOs) in radical pro-drop East-Asian languages. Focusing on Korean, we adopt as input the APNOs in the discourse-theoretically annotated corpus. Specifically, the corpus adopted is the one annotated by using the discourse features in the Centering Theory [5]. The current APNO resolution systems are supervised, thus relying on training data containing manually resolved APNOs. To eliminate the reliance on annotated data, we build an unsupervised classification model for the unsupervised Korean APNO resolution. We initially apply the unsupervised clustering algorithm and then train our unsupervised resolver on a set of training data from the corpus. After training, we use the resulting model to resolve Korean APNOs. The experiments show that our unsupervised ML model attains the same level of performance as the other supervised counterparts in efficiency when resolving Korean APNOs in the given corpus.
ISSN:1343-4500
1344-8994