Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Constituency Structure over Eojeol in Korean Treebanks

Jungyeul Park, Chulwoo Park

TL;DR

The paper addresses the representational challenge in Korean constituency annotation: should terminals be morphemes or eojeol? It argues for an eojeol-based constituency with morphology encoded in a separate non-constituent layer, enabling stable, cross-resource comparison and alignment with eojeol-based dependency resources. By normalizing Sejong and Penn Korean treebanks and mapping Kaist-like representations, it shows that a common eojeol backbone is feasible and beneficial. The proposed six-column annotation format, UPOS-labeled eojeol terminals, and parallel morphological layer provide a practical framework for interoperable corpus development and improved morphology-aware syntactic analysis in Korean.

Abstract

The design of Korean constituency treebanks raises a fundamental representational question concerning the choice of terminal units. Although Korean words are morphologically complex, treating morphemes as constituency terminals conflates word internal morphology with phrase level syntactic structure and creates mismatches with eojeol based dependency resources. This paper argues for an eojeol based constituency representation, with morphological segmentation and fine grained part of speech information encoded in a separate, non constituent layer. A comparative analysis shows that, under explicit normalization assumptions, the Sejong and Penn Korean treebanks can be treated as representationally equivalent at the eojeol based constituency level. Building on this result, we outline an eojeol based annotation scheme that preserves interpretable constituency and supports cross treebank comparison and constituency dependency conversion.

Constituency Structure over Eojeol in Korean Treebanks

TL;DR

The paper addresses the representational challenge in Korean constituency annotation: should terminals be morphemes or eojeol? It argues for an eojeol-based constituency with morphology encoded in a separate non-constituent layer, enabling stable, cross-resource comparison and alignment with eojeol-based dependency resources. By normalizing Sejong and Penn Korean treebanks and mapping Kaist-like representations, it shows that a common eojeol backbone is feasible and beneficial. The proposed six-column annotation format, UPOS-labeled eojeol terminals, and parallel morphological layer provide a practical framework for interoperable corpus development and improved morphology-aware syntactic analysis in Korean.

Abstract

The design of Korean constituency treebanks raises a fundamental representational question concerning the choice of terminal units. Although Korean words are morphologically complex, treating morphemes as constituency terminals conflates word internal morphology with phrase level syntactic structure and creates mismatches with eojeol based dependency resources. This paper argues for an eojeol based constituency representation, with morphological segmentation and fine grained part of speech information encoded in a separate, non constituent layer. A comparative analysis shows that, under explicit normalization assumptions, the Sejong and Penn Korean treebanks can be treated as representationally equivalent at the eojeol based constituency level. Building on this result, we outline an eojeol based annotation scheme that preserves interpretable constituency and supports cross treebank comparison and constituency dependency conversion.
Paper Structure (28 sections, 5 figures)

This paper contains 28 sections, 5 figures.

Table of Contents

  1. Introduction
  2. Background and conceptual distinctions
  3. Constituency, dependency, and the choice of terminal units
  4. Eojeol and morphemes in Korean grammar
  5. Existing Korean constituency treebanks
  6. The Sejong treebank
  7. The Penn Korean treebank
  8. The Kaist treebank
  9. Comparative summary and discussion
  10. Eojeol and constituency in Korean
  11. Why morpheme-based constituency terminals are problematic
  12. Toward an eojeol-based constituency representation with morphological annotation
  13. Representational objective of constituency
  14. Eojeol as the minimal syntactic terminal in Korean
  15. Separation of constituency and morphology by design
  16. ...and 13 more sections

Figures (5)

  • Figure 1: Constituency tree from the Sejong treebank
  • Figure 2: Constituency tree from the Penn Korean treebank
  • Figure 3: Constituency tree from the Kaist treebank
  • Figure 4: Six-column joint annotation format for eojeol-based constituency and morphological analysis
  • Figure 5: Eojeol-based constituency structure obtained from the Sejong treebank under eojeol-based normalization