Multilingual IR
IR: Information Retrieval
Overview
Multilingual information retrieval studies how to find relevant documents when queries and documents use different languages, scripts, or specialized vocabulary. Systems range from BM25 and hybrid sparse-dense search to multilingual embeddings, late-interaction models, translation, cross-lingual alignment, fine-tuning, and reranking. Evaluation spans benchmarks such as MIRACL and mMARCO as well as low-resource languages and specialized settings like medicine, law, and web search, often measuring retrieval for downstream question answering and RAG. Strong multilingual models can still lag behind language-specific systems, with performance varying by language and domain, so a central challenge is improving cross-language transfer without sacrificing accuracy, efficiency, or coverage for underrepresented languages.
Latest papers
53 papers7 papers in the last four weeks, up 75% on the four weeks before. 0.0% of all new papers.