MESSY STREETS：真实世界地址地理编码基准测试
原标题：MESSY STREETS: A Benchmark for Geocoding Real-World Addresses
AI 摘要
MESSY STREETS 是一个用于评估地理编码器在真实网络地址上性能的基准测试，基于 Web Data Commons 数据构建，并通过 OpenAddresses 或 OpenStreetMap 验证参考位置。研究发现，商业地理编码器在召回率上比开源系统高出最多 49 个百分点，主要差距源于对非规范地址的候选返回率；非规范表面形式可导致最多 25 个百分点的召回率损失。该基准测试首次对多种地理编码器在真实网络规模地址数据上的表现进行了公开比较，并指出标准化和预处理可显著缩小开源与商业系统之间的差距。
正文节选
MESSY STREETS: A Benchmark for Geocoding Real-World Addresses [experiments] Abstract. We introduce MESSY STREETS, a benchmark for evaluating geocoders on verbatim web addresses, with existence verification and controlled measurement of surface-form divergence. Unlike conventional benchmarks based on clean or synthetically perturbed addresses, MESSY STREETS contains addresses whose surface forms diverge from canonical representations and whose components may be missing, repeated, malformed, or in