The Art of Hierarchical Competing Patterns: Gaussian Process
Optimization of Hyphenation

D 2026

The Art of Hierarchical Competing Patterns: Gaussian Process Optimization of Hyphenation

SOJKA, Ondřej a Petr SOJKA

Základní údaje

Originální název

The Art of Hierarchical Competing Patterns: Gaussian Process Optimization of Hyphenation

Autoři

SOJKA, Ondřej a Petr SOJKA

Vydání

Budapest, Hungary, Findings of the Association for Computational Linguistics: EMNLP 2026, 18 s. 2026

Nakladatel

Association for Computational Linguistics

Další údaje

Jazyk

angličtina

Typ výsledku

Stať ve sborníku

Obor

10201 Computer sciences, information science, bioinformatics

Stát vydavatele

Spojené státy

Utajení

není předmětem státního či obchodního tajemství

Forma vydání

elektronická verze "online"

Odkazy

arXiv postprint, preprint, paper reproducibility repository

Označené pro přenos do RIV

Organizační jednotka

Fakulta informatiky

Příznaky

Mezinárodní význam, Recenzováno

Změněno: 9. 9. 2026 14:12, doc. RNDr. Petr Sojka, Ph.D.

Anotace

V originále

Hyphenation patterns remain a compact and widely deployed solution for word breaking in typesetting systems, text processors, and web rendering engines, but their generation still depends on manually tuned patgen program parameter profiles. We formulate patgen profile selection as a black-box hyperparameter optimization problem and evaluate Gaussian-process Bayesian optimization for this task. The search objective combines a precision-oriented $F_{1/7}$ score with an explicit trie size-accuracy tradeoff using a normalized trie-size penalty. We evaluate the method on 17 hyphenated word-list datasets covering 14 languages and multiple scripts. Against two strong hand-tuned profiles regenerated from the same 8/10 training split and evaluated on the same 1/10 held-out test split, the GP-optimized profiles improve $F_{1/7}$ and reduce trie size on all 17 datasets. The median optimized/baseline trie ratio is 0.348. A dataset-level sign test gives $p=7.6\cdot10^{-6}$; a separate budget-matched comparison on five representative datasets shows that model-based search improves over the best hand-tuned profile under the fixed comparison objective. The results show that model-based optimization can make pattern generation more reproducible and less dependent on expert trial-and-error while keeping the accuracy--compactness trade-off explicit.

Návaznosti

MUNI/A/1873/2025, interní kód MU

Název: Komplexní data a systémy s využitím umělé inteligence

Investor: Masarykova univerzita, Komplexní data a systémy s využitím umělé inteligence

Citovat

SOJKA, Ondřej a Petr SOJKA. The Art of Hierarchical Competing Patterns: Gaussian Process Optimization of Hyphenation. Online. In Findings of the Association for Computational Linguistics: EMNLP 2026. Budapest, Hungary: Association for Computational Linguistics, 2026, 18 s.

@inproceedings{2586799,
   author = {Sojka, Ondřej and Sojka, Petr},
   address = {Budapest, Hungary},
   booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2026},
   howpublished = {elektronická verze "online"},
   language = {eng},
   location = {Budapest, Hungary},
   publisher = {Association for Computational Linguistics},
   title = {The Art of Hierarchical Competing Patterns: Gaussian Process Optimization of Hyphenation},
   url = {https://arxiv.org/abs/2609.07638},
   year = {2026}
}

TY  - CONF
ID  - 2586799
AU  - Sojka, Ondřej - Sojka, Petr
PY  - 2026
TI  - The Art of Hierarchical Competing Patterns: Gaussian Process Optimization of Hyphenation
PB  - Association for Computational Linguistics
CY  - Budapest, Hungary
UR  - https://arxiv.org/abs/2609.07638
N2  - Hyphenation patterns remain a compact and widely deployed solution for word breaking in typesetting systems, text processors, and web rendering engines, but their generation still depends on manually tuned patgen program parameter profiles. We formulate patgen profile selection as a black-box hyperparameter optimization problem and evaluate Gaussian-process Bayesian optimization for this task. The search objective combines a precision-oriented $F_{1/7}$ score with an explicit trie size-accuracy tradeoff using a normalized trie-size penalty. We evaluate the method on 17 hyphenated word-list datasets covering 14 languages and multiple scripts. Against two strong hand-tuned profiles regenerated from the same 8/10 training split and evaluated on the same 1/10 held-out test split, the GP-optimized profiles improve $F_{1/7}$ and reduce trie size on all 17 datasets. The median optimized/baseline trie ratio is 0.348. A dataset-level sign test gives $p=7.6\cdot10^{-6}$; a separate budget-matched comparison on five representative datasets shows that model-based search improves over the best hand-tuned profile under the fixed comparison objective. The results show that model-based optimization can make pattern generation more reproducible and less dependent on expert trial-and-error while keeping the accuracy--compactness trade-off explicit.
ER  -

SOJKA, Ondřej a Petr SOJKA. The Art of Hierarchical Competing Patterns: Gaussian Process Optimization of Hyphenation. Online. In \textit{Findings of the Association for Computational Linguistics: EMNLP 2026}. Budapest, Hungary: Association for Computational Linguistics, 2026, 18 s.

Přehled o publikaci