Semantic-Symbolic Knowledge Consensus for Multilingual Question Answering
TL;DR
This paper proposes SeSyCo, a Semantic-Symbolic Knowledge Consensus framework, which leverages the semantic space to diverge monolingual queries into broad multilingual evidence, and subsequently utilize the symbolic space to eliminate language discrepancies, converging the gathered information into a robust consensus for precise SPARQL generation.
Abstract
Multilingual Question Answering (MQA) is primarily characterized by semantic-based and symbolic-based approaches, remain constrained by Language Confinement. Semantic methods, such as RAG and Agentic frameworks, suffer from linguistic bias that restricts retrieval and reasoning to the source language, while symbolic approaches struggle with cross-lingual schema alignment. We consider that integrating multilingual information from the semantic space into the symbolic space to formulate precise queries enables the effective utilization of global multilingual knowledge. In this paper, we propose SeSyCo, a Semantic-Symbolic Knowledge Consensus framework. Specifically, we leverage the semantic space to diverge monolingual queries into broad multilingual evidence, and subsequently utilize the symbolic space to eliminate language discrepancies, converging the gathered information into a robust consensus for precise SPARQL generation. Extensive experiments on the MLaKE dataset demonstrate that SeSyCo outperforms the strongest baseline by 10.7% in multi-hop settings, validating that establishing a robust multilingual consensus is essential for enhancing MQA performance. The collection is available at https://github.com/YuZhang9408/SeSyCo.