The RELX Dataset and Matching the Multilingual Blanks for Cross-Lingual Relation Classification

TL;DR

This paper introduces RELX, a new multilingual dataset and models for cross-lingual relation classification, addressing the scarcity of annotated data in low-resource languages and improving performance through distant supervision.

Contribution

The paper presents a new benchmark dataset RELX and a multilingual pretraining setup that enhances relation classification across multiple languages using distant supervision.

Findings

The new models outperform the baseline on the RELX benchmark.

RELX-Distant dataset contains hundreds of thousands of sentences with relations.

The approach effectively leverages Wikipedia and Wikidata for low-resource languages.

Abstract

Relation classification is one of the key topics in information extraction, which can be used to construct knowledge bases or to provide useful information for question answering. Current approaches for relation classification are mainly focused on the English language and require lots of training data with human annotations. Creating and annotating a large amount of training data for low-resource languages is impractical and expensive. To overcome this issue, we propose two cross-lingual relation classification models: a baseline model based on Multilingual BERT and a new multilingual pretraining setup, which significantly improves the baseline with distant supervision. For evaluation, we introduce a new public benchmark dataset for cross-lingual relation classification in English, French, German, Spanish, and Turkish, called RELX. We also provide the RELX-Distant dataset, which includes hundreds of thousands of sentences with relations from Wikipedia and Wikidata collected by distant supervision for these languages. Our code and data are available at: https://github.com/boun-tabi/RELX