AMR-to-Text Generation with Cache Transition Systems

TL;DR

This paper introduces a novel AMR-to-text generation approach that directly encodes graph structures and predicts parser actions to improve alignment and context understanding during sentence generation.

Contribution

It proposes a transition system-based model that encodes AMR graphs without linearization and predicts parser actions to enhance generation quality.

Findings

Two model variants with different action-word prediction strategies.

Direct graph encoding improves alignment over linearized methods.

Enhanced local context understanding in AMR-to-text generation.

Abstract

Text generation from AMR involves emitting sentences that reflect the meaning of their AMR annotations. Neural sequence-to-sequence models have successfully been used to decode strings from flattened graphs (e.g., using depth-first or random traversal). Such models often rely on attention-based decoders to map AMR node to English token sequences. Instead of linearizing AMR, we directly encode its graph structure and delegate traversal to the decoder. To enforce a sentence-aligned graph traversal and provide local graph context, we predict transition-based parser actions in addition to English words. We present two model variants: one generates parser actions prior to words, while the other interleaves actions with words.