EUVIMEDEUVIMED entdecken
Crossref

Dissecting the interplay between corpus properties, algorithm, and word segmentation performance

Cognitive Science Society 2025, Jun Ho Chai, Eon‐Suk Ko, Seongmin Mun

eScholarship (California Digital Library) · 18.06.2025

DOI: 10.48448/tces-n839 ↗

Kurz gefasst

Fachartikel in der Zeitschrift eScholarship (California Digital Library), erschienen am 18.06.2025. Verfasst von Cognitive Science Society 2025 und 3 weiteren Autor:innen.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study investigates how corpus-level properties in Korean child- and adult-directed speech shape word segmentation across four algorithms: Transitional Probability, Diphone-Based Segmentation, PUDDLE, and Adaptor Grammar. Utterance length consistently impacts segmentation, with shorter utterances improving performance, particularly for PUDDLE, DiBS, and AG. Word length affects transitional probability algorithms, while hapax legomena introduce challenges for forward TP and AG. Interjections negatively influence AG, but not the others, and larger corpus size benefits PUDDLE. Register effects are limited, with forward TP and PUDDLE performing better on child-directed speech. These patterns highlight algorithm-specific sensitivities, with utterance length emerging as the most consistent factor. Our findings underscore the importance of considering both input properties and algorithm design when studying word segmentation in Korean. Future work should explore cross-linguistic comparisons, larger balanced corpora, and the role of multimodal cues in segmentation.

Schlagwörter

word segmentationKorean languagechild-directed speechadult-directed speechtransition probabilitiespuddlesutterance lengthword lengthhapax legomenainterjections

Zitieren

APA

2025, C. S. S., Chai, J. H., Ko, E., & Mun, S. (2025). Dissecting the interplay between corpus properties, algorithm, and word segmentation performance. eScholarship (California Digital Library). https://doi.org/10.48448/tces-n839

Vancouver

2025 CSS, Chai JH, Ko E, Mun S. Dissecting the interplay between corpus properties, algorithm, and word segmentation performance. eScholarship (California Digital Library). 2025. doi:10.48448/tces-n839

BibTeX
@article{2025,
  author = {2025, Cognitive Science Society and Chai, Jun Ho and Ko, Eon‐Suk and Mun, Seongmin},
  title = {Dissecting the interplay between corpus properties, algorithm, and word segmentation performance},
  journal = {eScholarship (California Digital Library)},
  year = {2025},
  doi = {10.48448/tces-n839}
}
Kostenlos starten →