Algorithms for Deterministic Incremental Dependency Parsing

6.2 Parsing Accuracy

1. What were the main accuracy results across 13 languages?
   Averages were very close, within 0.58 LAS points: 79.77 for non-projective list-based (best) down to 79.19 for strictly projective list-based. Non-projective and pseudo-projective parsers significantly beat strictly projective ones, and the non-projective parser’s gain grew with the proportion of non-projective arcs in the data (Pearson’s r = 0.815).
   • Nivre, 2008, Algorithms for Deterministic Incremental Dependency Parsing (Computational Linguistics), §6.2, pp. 541–546
2. How do non-projective and pseudo-projective parsing compare?
   Both help when non-projective arcs are common, but non-projective parsing is more stable: it seldom or never hurts, while pseudo-projective parsing can hurt when training data or non-projective arcs are scarce, partly because extra labels reduce label accuracy.
   • Nivre, 2008, Algorithms for Deterministic Incremental Dependency Parsing (Computational Linguistics), §6.2, pp. 541–546
3. Why did arc-standard parsing do best on Arabic, Czech and Slovene?
   Their Prague-style annotation allows several root dependents with different labels (e.g. PRED), which must be attached by a RIGHT-ARC with the right label at the right time. Arc-eager attaches right dependents early and had much lower precision and recall on PRED than arc-standard.
   • Nivre, 2008, Algorithms for Deterministic Incremental Dependency Parsing (Computational Linguistics), §6.2, pp. 541–546