A Fast Linkage Detection Scheme for Multi （多的快速链接检测计划）.pdfVIP

A Fast Linkage Detection Scheme for Multi-Source Information Integration Akiko Aizawa Keizo Oyama National Intsitute of Informatics / National Intsitute of Informatics / The Graduate University for Advanced Studies The Graduate University for Advanced Studies 2-1-2 Hitotsubashi, Chiyoda-ku, 2-1-2 Hitotsubashi, Chiyoda-ku, Tokyo 101-8430, Japan Tokyo 101-8430, Japan aizawa@nii.ac.jp oyama@nii.ac.jp Abstract [24, 2] or deduplication [27], which emphasizes the elimi- nation of duplications in a database; record matching [31] or Record linkage refers to techniques for identify- approximate string join [10], which is an operation to inte- ing records associated with the same real-world enti- grate two separate tables; entity-name clustering and match- ties. Record linkage is not only crucial in integrating ing [4], object consolidation [23] or proper noun corefer- multi-source databases that have been generated inde- ence analysis [22], which concerns the coreference determi- pendently, but is also considered to be one of the key nation problem in text processing; and entity identification issues in integrating heterogeneous Web resources. How- [18], object identification [30], entity matching [5] or entity ever, when targeting large-scale data, the cost of enu- reconciliation [6], which concerns h