Character-level seq2seq models for Japanese past-tense inflection show 75-80% of residual errors tied to gemination, especially in e-stem verbs, with consistent patterns across architectures that motivate orthography-aware evaluation.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Mind Your Moras: Orthography-Aware Error Analysis of Neural Japanese Morphological Generation
Character-level seq2seq models for Japanese past-tense inflection show 75-80% of residual errors tied to gemination, especially in e-stem verbs, with consistent patterns across architectures that motivate orthography-aware evaluation.