Parsing Universal Dependencies without training

TL;DR

UDP is a novel training-free parser for Universal Dependencies that uses PageRank and simple rules, providing a robust, unsupervised, and linguistically sound alternative to traditional cross-lingual parsing methods.

Contribution

It introduces UDP, the first training-free UD parser, combining PageRank with head attachment rules, and demonstrates its effectiveness across languages and domains.

Findings

Competitive with delexicalized transfer systems

Robust to domain changes across languages

Requires no training or extensive parameters

Abstract

We propose UDP, the first training-free parser for Universal Dependencies (UD). Our algorithm is based on PageRank and a small set of head attachment rules. It features two-step decoding to guarantee that function words are attached as leaf nodes. The parser requires no training, and it is competitive with a delexicalized transfer system. UDP offers a linguistically sound unsupervised alternative to cross-lingual parsing for UD, which can be used as a baseline for such systems. The parser has very few parameters and is distinctly robust to domain change across languages.