Constructive Type-Logical Supertagging with Self-Attention Networks
This work addresses grammar induction for computational linguistics, offering a novel approach that improves generalization beyond closed-world assumptions, though it appears incremental in applying self-attention to a specific grammar task.
The paper tackles grammar induction by proposing a self-attention network-based supertagger for a refined type-logical grammar, achieving high overall type accuracy and learning the grammar's syntax and semantics, which enhances generalization by handling sparse word types and constructing unseen complex types.
We propose a novel application of self-attention networks towards grammar induction. We present an attention-based supertagger for a refined type-logical grammar, trained on constructing types inductively. In addition to achieving a high overall type accuracy, our model is able to learn the syntax of the grammar's type system along with its denotational semantics. This lifts the closed world assumption commonly made by lexicalized grammar supertaggers, greatly enhancing its generalization potential. This is evidenced both by its adequate accuracy over sparse word types and its ability to correctly construct complex types never seen during training, which, to the best of our knowledge, was as of yet unaccomplished.