Language Bias in Information Retrieval: The Nature of the Beast and Mitigation Methods
This addresses fairness for users in multilingual information retrieval, but it appears incremental as it builds on existing neural methods.
The paper tackled language bias in multilingual information retrieval by evaluating fairness across methods and introducing LaKDA, a novel loss that effectively enhanced fairness, though specific numerical improvements were not detailed.
Language fairness in multilingual information retrieval (MLIR) systems is crucial for ensuring equitable access to information across diverse languages. This paper sheds light on the issue, based on the assumption that queries in different languages, but with identical semantics, should yield equivalent ranking lists when retrieving on the same multilingual documents. We evaluate the degree of fairness using both traditional retrieval methods, and a DPR neural ranker based on mBERT and XLM-R. Additionally, we introduce `LaKDA', a novel loss designed to mitigate language biases in neural MLIR approaches. Our analysis exposes intrinsic language biases in current MLIR technologies, with notable disparities across the retrieval methods, and the effectiveness of LaKDA in enhancing language fairness.