How NameSearch® Works - Intelligent Scoring

NameSearch® comparison algorithms determine numeric values that represent the likelihood of a match. Two entities are passed to the comparison routine and a score or array of scores are returned.

There are many ways of calculating a score. William Jefferson Clinton versus Bill Clinton would yield as scores; 100, 50, 66 and 33. The basic approach is to divide the number of tokens by the number of matches and multiply that figure by 100. In the previous example the number of tokens is either two or three words. Bill Clinton has two tokens and William Jefferson Clinton has three. The only word that matches exactly is Clinton. Dividing the number of tokens by the number of matches and multiplying by 100 we get (1/2 * 100 = 50) and (1/3 * 100 = 33). Alternatively, Bill and William are used interchangeably and can be considered to match. In this instance the scores would be (2/2 *100 = 100) and (2/3 * 100 = 66).

The score or scores that are used to determine the likelihood of a match, depend on the degree of accuracy required by your system. Scoring is essential in on-line applications, where the number of records returned is too great for a person to scan or batch utilities, where decisions based on the likelihood of a match invoke automated processes.

ALFACOMP Comparison Algorithm
This is used to compare fields containing multi word strings. The ALFACOMP routine is based solely on a heuristic algorithm and is not dependent on rulebase expertise.

DATESCR Comparison Algorithm
The date compare routine is used to intelligently determine the similarity of two dates. The routine uses the strength of NameSearch® to arrive at its results. The date comparison routine interrogates two dates by parsing them into their components. The month and day represent one component, which accounts for half the score; the year field represents the other half.
The first part of the scoring routine is the parsing of information. The majority of this work is accomplished through the use of NameSearch®’s sanitization and word recognition routines. These routines use the date service, supplied with the NameSearch® software, to arrive at its results.
After the dates have been parsed and standardized, the date comparison back-end calculates a score between 0 and 100.
The DATESCR comparison routine uses rulebase expertise in order to arrive at its’ results. For example, July 28, 1965 compared to 7/28/66 would yield a score of 100 in this matter the DATESCR routine overcomes problems due to inconsistency in date format. The routine also accepts several parameters which will dictate the penalty for mis-matches based on the year. By increasing these settings the score can be made more tolerant. For example, if you want all dates that correspond to July 28, 1965 + or - two years, you would set the year range to 2. If you wish it to be + or - five years this would mean your year range would be set to 5. In this manner NameSearch® gives you the ability to widen or narrow the range of dates being returned given the month a day agrees.

NUMCOMP Comparison Algorithm
The number comparison routine compares two alpha-numeric fields. The routine was originally designed for numeric character comparisons but works well on alpha-numeric character strings. For example this routine is well suited for Social Security number comparisons.
The numcmp function compares two strings, character by character. It returns four scores. Each score represents a different way of calculating the number of characters, that correctly compared against the number of characters available for comparison.

COMP, COMP1, COMP2 Comparison Algorithm
These are NameSearch®’s comparison routines used for scoring names and addresses. These routines utilize NameSearch®’s rulebase expertise and phonetic tokenization to determine scores. Comp was the original comparison routine released with version 1 of the NameSearch® product. COMP1 was introduced in Version 2.0 of the NameSearch product in order to provide a more representative score. In Version 2.5 of the NameSearch product COMP2 was added. This routine uses ALFACOMP, an advanced heuristic algorithm, to arrive at its results.


MultiComp Comparison Algorithm
MultiComp routine allows evaluating multiple fields and delivering a combined, weighted score. For example, users may want to compare personal names, corporate names and addresses, and receive a single score. This score is created based on a field definition string that contains the methodology used for deriving and weighing scores.
MultiComp provides the capability of comparing fields with different comparison routines, setting up scoring scheme masks, customizing threshold scores, as well as weighting attributes and score types.

How NameSearch® works