Applying Multilingual Models to Question Answering (QA)

TL;DR

This paper evaluates monolingual and multilingual models on question-answering tasks across English, Finnish, and Japanese, focusing on answerability and answer span identification, including cross-language zero-shot transfer with Multilingual BERT.

Contribution

It introduces models for answerability and answer span detection in multilingual QA and assesses Multilingual BERT's zero-shot capabilities across languages.

Findings

Multilingual models perform well on QA tasks across diverse languages.

Multilingual BERT shows promising zero-shot transfer performance.

Models effectively identify answer spans and determine answerability in multiple languages.

Abstract

We study the performance of monolingual and multilingual language models on the task of question-answering (QA) on three diverse languages: English, Finnish and Japanese. We develop models for the tasks of (1) determining if a question is answerable given the context and (2) identifying the answer texts within the context using IOB tagging. Furthermore, we attempt to evaluate the effectiveness of a pre-trained multilingual encoder (Multilingual BERT) on cross-language zero-shot learning for both the answerability and IOB sequence classifiers.