Much of the research related to information retrieval focuses on finding as many relevant pieces of data about a topic as possible. With respect to identity matching, this approach would find as many possible variants of a search name as possible. This works fine for some applications, but there are also times when the requirement is for a best fit match. In this case, we expect that a single name has zero or one matches in another data set. Many systems accomplish this by use of a hard key, such as driver's license number or social security number. This paper presents SQL based soft matching to get a best fit match. This is valuable when one or more of the data sets being compared has data quality problems or lacks complete and reliable hard keys.
展开▼