Cross-lingual Extended Named Entity Classification of Wikipedia Articles

TL;DR

This paper presents a cross-lingual approach for Wikipedia article classification, leveraging multilingual representations and a three-stage training process, achieving high accuracy across 30 languages.

Contribution

It introduces a novel three-stage method combining multilingual pre-training, monolingual fine-tuning, and cross-lingual voting for improved Wikipedia article classification.

Findings

Achieved top scores in 25 of 30 languages.

Small accuracy gaps in the remaining five languages.

Effective cross-lingual representations for page classification.

Abstract

The FPT.AI team participated in the SHINRA2020-ML subtask of the NTCIR-15 SHINRA task. This paper describes our method to solving the problem and discusses the official results. Our method focuses on learning cross-lingual representations, both on the word level and document level for page classification. We propose a three-stage approach including multilingual model pre-training, monolingual model fine-tuning and cross-lingual voting. Our system is able to achieve the best scores for 25 out of 30 languages; and its accuracy gaps to the best performing systems of the other five languages are relatively small.