Unsupervised Search-based Structured Prediction

TL;DR

This paper adapts the search-based structured prediction algorithm 'Searn' for unsupervised learning, demonstrating high-quality parsing, connections to EM, and a semi-supervised extension using predict-self ideas.

Contribution

It introduces a novel adaptation of Searn for unsupervised learning, linking it to EM and enabling semi-supervised learning with high-quality results.

Findings

High-quality unsupervised shift-reduce parsing achieved

Connection established between Searn and EM algorithms

Effective semi-supervised extension demonstrated

Abstract

We describe an adaptation and application of a search-based structured prediction algorithm "Searn" to unsupervised learning problems. We show that it is possible to reduce unsupervised learning to supervised learning and demonstrate a high-quality unsupervised shift-reduce parsing model. We additionally show a close connection between unsupervised Searn and expectation maximization. Finally, we demonstrate the efficacy of a semi-supervised extension. The key idea that enables this is an application of the predict-self idea for unsupervised learning.