Improving Korean-English Cross-Lingual Retrieval: A Data-Centric Study of Language Composition and Model Merging

TL;DR

This study investigates how the composition of training data affects Korean-English cross-lingual retrieval and shows that model merging can balance performance trade-offs between CLIR and mono-lingual IR.

Contribution

It systematically analyzes the impact of language data composition on retrieval performance and introduces model merging as a strategy to optimize cross-lingual and mono-lingual IR.

Findings

Language composition significantly influences IR performance.

CLIR improves with specific language pairs, mono-lingual IR declines.

Model merging mitigates performance trade-offs.

Abstract

With the increasing utilization of multilingual text information, Cross-Lingual Information Retrieval (CLIR) has become a crucial research area. However, the impact of training data composition on both CLIR and Mono-Lingual Information Retrieval (IR) performance remains under-explored. To systematically investigate this data-centric aspect, we construct linguistically parallel Korean-English datasets and train retrieval models with various language combinations. Our experiments reveal that the language composition of training data significantly influences IR performance, exhibiting important inter-lingual correlations: CLIR performance improves with specific language pairs, while Mono-Lingual IR performance declines. Our work demonstrates that Model Merging can effectively mitigate this trade-off, achieving strong CLIR results while preserving Mono-Lingual IR capabilities. Our findings underscore the effects of linguistic configuration of training data on both CLIR and Mono-Lingual IR, and present Model Merging as a viable strategy to optimize performance across these tasks.