Data, naturally, is fuzzy. There are many reasons for that; mistakes in typing, using abbreviations, and so on. The following screenshot (sourced from Microsoft) shows an example of two records for the same person:

From the human point of view, both the records shown in the preceding screenshot are for the same person; it just has some abbreviations and different string formats. But from the computer's point of view, these records are different; or, in the other words, they are not exactly similar.

The data matching component of DQS works with a similarity threshold between domain values. The data steward can create matching policies in the Knowledge Base. Each matching policy contains one or more matching rules. Matching rules define how records will match each other. In the matching rules, the type of similarity can be defined as prerequisite, exact match, or similar. Matching rules can be tuned incrementally with the incoming data with the data steward's supervision.

There are four...