Papers › Self-Attention Networks Can Process Bounded Hierarchical Languages

Self-Attention Networks Can Process Bounded Hierarchical Languages

24 May 2021ACL 2021 5arXiv:2105.11115archive 2025-07-28

Shunyu Yao, Binghui Peng, Christos Papadimitriou, Karthik Narasimhan

Despite their impressive performance in NLP, self-attention networks were recently proved to be limited for processing formal languages with hierarchical structure, such as 𝖣𝗒𝖼𝗄ₖ, the language consisting of well-nested parentheses of k types. This suggested that natural language can be approximated well with models that are too weak for formal languages, or that the role of hierarchy and recursion in natural language might be limited. We qualify this implication by proving that self-attention networks can process 𝖣𝗒𝖼𝗄_(k, D), the subset of 𝖣𝗒𝖼𝗄ₖ with depth bounded by D, which arguably better captures the bounded hierarchical structure of natural language. Specifically, we construct a hard-attention network with D+1 layers and O(logk) memory size (per token per layer) that recognizes 𝖣𝗒𝖼𝗄_(k, D), and a soft-attention network with two layers and O(logk) memory size that generates 𝖣𝗒𝖼𝗄_(k, D). Experiments show that self-attention networks trained on 𝖣𝗒𝖼𝗄_(k, D) generalize to longer inputs with near-perfect accuracy, and also verify the theoretical memory advantage of self-attention networks over recurrent networks.

Code

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Results from the paper archive 2025-07-28

No leaderboard rows for this paper in the archive.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections