Errata for Data Matching - Concepts and Techniques for Record Linkage, Entity Resolution, and Duplicate Detection

Last update: 5 May 2021.

Thanks to everybody who has reported an error.

Please report any further errors you notice to Peter Christen

Page 80, Figure 4.5: In the sub-figure in top-right (Blocks A: GivenName), the third row should have a Soundex encoding of `p360' instead of `p630'.
Page 80, again in Figure 4.5: In the sub-figure Candidate record pairs from GivenName, the pair of record identifiers should be (a3,b2) instead of (a3,b3).
Page 81, Equation (4.2) is wrong. We have n records and b blocks. In each block, we compare n/b * (n/b-1)/2 record pairs, so c = b * n/b*(n/b-1)/2 = n/2 * (n/b-1).
Page 91: The descriptions of Document Frequency and Term Frequency need to be flipped. This means the sentence: "The Document Frequency (DF) is the count of how many times a token appears in a BKV in a certain record" should be: "The Term Frequency (TF) is the count of how many times a token appears in a BKV in a certain record", and: "The Term Frequency (TF), is the count of how many records in a database contain a certain token" should be: "The Document Frequency (DF), is the count of how many records in a database contain a certain token". Equation (4.8) then needs to be changed accordingly, with tf(n) replaced by df(n).
Page 96, last paragraph, the sentence: "The idea is to build a relationship graph where records are nodes and relationships between records are vertices." should say: "The idea is to build a relationship graph where records are nodes, and relationships between records are edges."
Section 5.4, page 107: In the example at the bottom of the page, the numerical values in the sim_jaccard equation should be:
sim_jaccard(`gail', `gayle') = 1 / (3 + 4 - 1) = 1 / 6 = 0.16667,
instead of:
sim_jaccard(`gail', `gayle') = 1 / (3 + 3 - 1) = 1 / 5 = 0.2.
Section 10.2, page 231: Another freely available data matching system which unfortunately was missed is Duke. For details see:
http://code.google.com/p/duke/
Duke is a practical data matching system developed in Norway, with the aim to be easy to embedded into other software.