Search papers, labs, and topics across Lattice.
This paper redesigns the symbolic backend of a parsing system using CCG directed types to enhance structural generalization in language parsing tasks. By implementing a deterministic CKY algorithm with a single linear decoder and a modest parameter count, the new system achieves a significant performance boost over the previous state-of-the-art AM-Parser, particularly in directional categories. The results indicate that while directional representations improve performance in specific areas, the overall bottleneck shifts to the neural layer, suggesting that encoder quality plays a critical role in parsing efficacy.
Directional representations in parsing systems can lead to a dramatic 29.9 percentage point improvement in performance on position-shift categories, shifting the bottleneck from symbolic to neural layers.
Several SLOG test categories explicitly involve directional distinctions (modifier position shifts, argument extraction positions), yet AM-Parser, the previous SOTA, uses an AM algebra whose operations do not encode direction. We redesign the symbolic backend around CCG directed types (deterministic CKY + single linear decoder, 30K learnable parameters). Under the same BERT-base encoder, the system achieves 75.9$\pm$6.4% LF exact match, surpassing AM-Parser (70.8$\pm$4.3%). Per SLOG's own category groupings, gains are highly directional: the CCG system outperforms AM-Parser on all 5 position-shift categories (+29.9pp), while AM-Parser outperforms on all 6 recursive-depth categories. Replacing the encoder with DeBERTa-v3-large yields 90.7$\pm$4.9%, with the largest encoder gains in recursive-depth categories, complementary to directionality's gains. Directional representations shift the bottleneck from the symbolic layer (AM-Parser's 0% category ceiling) to the neural layer, which improves with encoder upgrades.
