Cross-lingual, Character-Level Neural Morphological Tagging

TL;DR

This paper introduces a transfer learning approach using character-level neural models to improve morphological tagging in low-resource languages by leveraging related high-resource languages, achieving significant accuracy gains.

Contribution

It presents a novel joint training scheme for character-level neural taggers across multiple related languages, enabling effective knowledge transfer and improved performance in low-resource settings.

Findings

Up to 30% accuracy improvement over monolingual models.

Joint character representations facilitate cross-lingual transfer.

Effective for low-resource languages with related high-resource counterparts.

Abstract

Even for common NLP tasks, sufficient supervision is not available in many languages -- morphological tagging is no exception. In the work presented here, we explore a transfer learning scheme, whereby we train character-level recurrent neural taggers to predict morphological taggings for high-resource languages and low-resource languages together. Learning joint character representations among multiple related languages successfully enables knowledge transfer from the high-resource languages to the low-resource ones, improving accuracy by up to 30% over a monolingual model.