Better heads do not guarantee better binarized constituency parsing

2026-05-27

AI Summary

This study investigates whether dependency-induced head information can enhance the performance of binarized constituency parsing in punctuation-sensitive scenarios. By comparing learned heads against rule-based heads in punctuation-aware tree binarization and evaluating downstream parsing quality after debinarization, the authors find that more accurate head prediction does not necessarily yield better parsing performance, thereby challenging the assumption that linguistically motivated heads serve as effective control signals for binarization. Experimental results show modest overall improvements on the Chinese Treebank (CTB), yet the approach exhibits unstable performance in punctuation-sensitive macro-averaged F1 scores and cross-treebank transfer tasks, indicating that learned heads are less robust than their rule-based counterparts.

Abstract

We revisit punctuation-aware tree binarization for constituency parsing and ask whether dependency-induced headedness improves binary parser supervision. Although learned heads substantially outperform rule-based heads in intrinsic head prediction, they do not yield consistent parsing gains after debinarization. In particular, punctuation-conditioned evaluation shows that learned headedness underperforms rule-based binarization in macro-average punctuation-sensitive $F_1$, despite a small overall gain on CTB. Similar instability appears under cross-treebank transfer. These results suggest that \ycc{linguistically grounded} headedness is not necessarily parser-optimal when used as a binarization control signal. The paper presents a negative result: better head prediction does not imply better punctuation-sensitive constituency parsing.

Problem

• constituency parsing
• binarization
• headedness
• punctuation-sensitive evaluation
• cross-treebank transfer

Innovation

• binarized constituency parsing
• headedness
• punctuation-aware evaluation
• dependency-induced binarization
• negative result