Enhancing Multilingual Accessibility of Question Answering over Knowledge Graphs

Abstract

There are more than 7000 languages spoken in the world today. Yet, English dominates in many research communities, in particular in the field of Knowledge Graph Question Answering (KGQA). The goal of a KGQA system is to provide natural-language access to a knowledge graph. While many research works aim to achieve the best possible QA quality over English benchmarks, only a small portion of them focuses on providing these systems in a way that different user groups (e.g., speakers of different languages) may use them with the same efficiency (i.e., accessibility). To address this research gap, we investigate the multilingual aspect of the accessibility, which enables speakers of different languages (including low-resource and endangered languages) to interact with KGQA systems with the same efficiency.