SPINDLE: Spinning Raw Text into Lambda Terms with Graph Attention

TL;DR

SPINDLE is an open-source Python parser for Dutch that converts raw text into lambda calculus programs, integrating recent advances in multimodal grammar, graph convolutions, and parallel proof search for accurate meaning representation.

Contribution

It introduces a novel parser combining multimodal type-logical grammar, graph neural networks, and Sinkhorn-based proof search, advancing the state-of-the-art in Dutch natural language understanding.

Findings

High accuracy in Dutch semantic parsing

Efficient proof search with Sinkhorn iterations

Open source implementation with visualization tools

Abstract

This paper describes SPINDLE - an open source Python module implementing an efficient and accurate parser for written Dutch that transforms raw text input to programs for meaning composition, expressed as {\lambda} terms. The parser integrates a number of breakthrough advances made in recent years. Its output consists of hi-res derivations of a multimodal type-logical grammar, capturing two orthogonal axes of syntax, namely deep function-argument structures and dependency relations. These are produced by three interdependent systems: a static type-checker asserting the well-formedness of grammatical analyses, a state-of-the-art, structurally-aware supertagger based on heterogeneous graph convolutions, and a massively parallel proof search component based on Sinkhorn iterations. Packed in the software are also handy utilities and extras for proof visualization and inference, intended to facilitate end-user utilization.