Better heads do not guarantee better binarized constituency parsing
For researchers in constituency parsing, this paper presents a negative result showing that linguistically grounded headedness is not necessarily optimal for parser binarization.
The paper investigates whether dependency-induced headedness improves binary parser supervision for constituency parsing. It finds that learned heads outperform rule-based heads in head prediction but do not yield consistent parsing gains, with rule-based binarization often performing better in punctuation-sensitive F1.
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.