MultiLingual Information Retrieval
Multilingual information retrieval (MLIR) aims to build systems that effectively search and rank documents across multiple languages, overcoming the limitations of monolingual search engines. Current research focuses on developing robust multilingual models, often leveraging transformer architectures and contrastive learning, to address challenges like handling low-resource languages, reducing language bias, and improving efficiency in both training and inference. These advancements have significant implications for broader information access, particularly for under-resourced languages and communities, and are driving progress in areas like cross-lingual question answering and multilingual document understanding.
Papers
Language Fairness in Multilingual Information Retrieval
Eugene Yang, Thomas Jänich, James Mayfield, Dawn Lawrie
Distillation for Multilingual Information Retrieval
Eugene Yang, Dawn Lawrie, James Mayfield
PLAID SHIRTTT for Large-Scale Streaming Dense Retrieval
Dawn Lawrie, Efsun Kayi, Eugene Yang, James Mayfield, Douglas W. Oard