arXiv: domain-aware distribution alignment for budgeted entity matching
arXiv·low signal
Pulsone, Goren, and Shraga study entity matching — comparing records across sources in data-integration pipelines — under a constrained labeling budget, proposing domain-aware distribution alignment to spend scarce labels more effectively. The method targets the practical bottleneck of expensive human annotation in record-linkage and deduplication work. It's a targeted contribution for teams building data-integration or RAG-ingestion pipelines where label budgets dominate cost.