Understanding the Sun Match Engine

Sun Match Engine Matching Configuration

The matching configuration files define how the Sun Match Engine processes records to assign matching probability weights, allowing the master index application to identify matches, potential duplicates, and non-matches. These files consist of two configurable files, the match configuration file and the match constants file. Together these files define additional logic for the Sun Match Engine to use when determining the matching probability between two records. A third file, the internal match constants file, is read-only and used internally by the match engine. It defines each comparison function and the comparison options.

The matching configuration files are very flexible, allowing you to customize the matching logic according to the type of data stored in the master index application and for the record matching requirements of your business. The matching configuration files are stored in the master index project and appear as nodes in the Match Engine node of the project. The Sun Match Engine typically standardizes the data prior to matching, so the match process is performed against the standardized data.

The matching configuration files for the Sun Match Engine must follow certain rules for formatting and interdependencies. The following topics provide an overview of the two matching configuration files provided, the architecture of those files, and formatting descriptions. They also include an overview of comparison functions used in the match configuration file.

The Sun Match Engine Match Configuration File

The match configuration file, matchConfigFile.cfg, contains the matching logic for each field on which matching is performed. By default, this file defines the matching logic for the three primary data types (person names, business names, and addresses), and can also handle generic data types, such as dates, numbers, social security numbers, and characters.

The match configuration file defines matching logic for each field on which matching is performed. The Sun Match Engine provides several comparison functions that you can call in this file to fine-tune the match process. Comparison functions contain the logic to compare different types of data in very specific ways in order to arrive at a match weight for each field. These functions allow you to define how matching is performed for different data types and can be used in conjunction with either matching and unmatching probabilities or agreement and disagreement weight ranges for each field. This file also defines how to handle missing fields.

The following topics describe the format of the configuration file and provide an overview of the predefined comparison functions:

Sun Match Engine Match Configuration File Format

The match configuration file is divided into two sections. The first section consists of one line that indicates the matching probability type. The second section consists of the matching rules to use for each match field.

Match Configuration File Sample

Following is an excerpt from the default match configuration file. This excerpt illustrates the components that are described in the following sections.


ProbabilityType            1

FirstName              15  0   uf    0.99  0.001   15  -5
LastName               15  0   ul    0.99  0.001   15  -5
String                 25  0   ua    0.99  0.001   10  -5
DateDays               20  0   dD    0.99  0.001   10  -10 y 15      30
DateMonths             20  0   dM    0.99  0.001   10  -10 n
DateHours              20  0   dH    0.99  0.001   10  -10 y 30      60
DateMinutes            20  0   dm    0.99  0.001   10  -10 y 300 600
DateSeconds            20  0   ds    0.99  0.001   10  -10 y 75      60
Numeric                15  0   n     0.99  0.001   10  -10 y 8
Integer                15  0   nI    0.99  0.001   10  -10 n
Real                   15  0   nR    0.99  0.001   10  -10 n
Char                   1   0   c     0.99  0.001   5   -5
pro                    15  0   p     0.99  0.001   10  -10 20 5 5

Probability Type

The first line of the match configuration file defines the probability type to use for matching. Specify “0” (zero) to use m-probabilities and u-probabilities to determine a field’s match weight; specify “1” (one) to use agreement and disagreement weight ranges. If the probability type is set to use agreement and disagreement weight ranges, the m-prob and u-prob columns in the matching rules section are ignored. Likewise, if the probability type is set to use m-probabilities and u-probabilities, the agreement-weight and disagreement-weight columns in the matching rules section are ignored. The default is to use agreement and disagreement weight ranges because they are more intuitive.

Matching Rules

The section after the first line of the match configuration file contains match field rows, with each row defining how a certain data type or field will be matched. The syntax for this section is:

match-type size null-field function m-prob u-prob agreement-weight disagreement-weight parameters

Table 1 describes each element in a match field row.

Table 1 Match Configuration File Columns

Column Number 

Column Name 

Description 

match-type 

A value that indicates to the Sun Match Engine how each field should be weighted. Each field included in the match string (the MatchingConfig section of the Match Field file) must have a match type corresponding to a value in this column.

size 

The number of characters in the field on which matching is performed, beginning with the first character. For example, to match on only the first four characters in a 10-digit field, the value of this column should be “4”. 

null-field 

An index that specifies how to calculate the total weight for null fields or fields that only contain spaces. You can specify any of the following values:

  • 0 - (zero) If one or both fields are empty, the weight used for the field is 0 (zero).

  • 1 - (one) If both fields are empty, the agreement weight is used; if only one field is empty, the disagreement weight is used.

  • a# - An “a” followed by a number specifies to use the agreement weight if both fields are empty. The agreement weight is divided by the number following the “a” to obtain the match weight for that field. If no number is specified, the default is “2”. You can specify any number from 1 through 10.

  • d# - A “d” followed by a number specifies to use the disagreement weight if only one field is empty. The disagreement weight is divided by the number following the “d” to obtain the match weight for the field. If no number is specified, the default is “2”. You can specify any number from 1 through 10.


Note –

In the above descriptions, the agreement and disagreement weights are either specified in this file or calculated using a logarithmic formula based on the m and u-probabilities (depending on the probability type).


function 

The type of comparison to perform when weighting the field. For information about the available comparison functions, see Match Configuration Comparison Functions for Sun Match Engine (Repository).

m-prob 

The initial probability that the specified field in two records will match if the records match. The probability is a double value between 0 and 1, and can have up to 16 decimal points. 

u-prob 

The initial probability that the specified field in two records will match if the records do not match. The probability is a double value between 0 and 1, and can have up to 16 decimal points. 

agreement-weight 

The matching weight to be assigned to a field given that the fields match between two records. This number can be between 0 and 100 and can have up to 16 decimal points. It represents the maximum match weight for a field. 

disagreement-weight 

The matching weight to be assigned to a field given that the fields do not match between two records. This number can be between 0 and -100 and can have up to 16 decimal points. It represents the minimum match weight for a field. 

parameters 

The parameters that correspond to the comparison function specified in column 4. Some comparison functions do not take any parameters and some take multiple parameters. For additional information about parameters, see Match Configuration Comparison Functions for Sun Match Engine (Repository).

Sun Match Engine Matching Comparison Functions

Match field comparison functions, or comparators, compare the values of a field in two records to determine whether the fields match. The fields are then assigned a matching weight based on the results of the comparison function. You can use several different types of comparison functions in the match configuration file to define how the Sun Match Engine should match the fields in the match string. The Sun Match Engine provides several options to use with each function.

The following table summarizes each comparison function. A complete reference of the comparison functions and their parameters is included in Match Configuration Comparison Functions for Sun Match Engine (Repository).

Table 2 Comparison Functions

Comparison Function 

Name 

Description 

b1 

Bigram String Comparator 

Based on the Bigram algorithm, this function compares two strings using all combinations of two consecutive characters and returns the total number of combinations that are the same.

b2 

Advanced Bigram String Comparator 

Similar to the standard Bigram comparison function (b1), but allows for character transpositions.

Generic String Comparator 

Based on the Jaro algorithm, this function compares two strings taking into account uncertainty factors, such as string length, transpositions, and characters in common.

ua 

Advanced Generic String Comparator 

Based on the Jaro algorithm with variants of Winkler/Lynch and McLaughlin, this function is similar to the generic string comparator (u), but increases the agreement weight if the initial characters of each string are exact matches. This comparison function takes into account key punch and visual memory errors.

uf 

Simplified String Comparator - FirstName 

Based on the generic string comparator (u), this function is designed to specifically weight first name values. The string is analyzed and the weight adjusted based on statistical data. 

ul 

Simplified String Comparator - LastName 

Based on the generic string comparator (u), this function is designed to specifically weight last name values. The string is analyzed and the weight adjusted based on statistical data.

un 

Simplified String Comparator - House Numbers 

Based on the generic string comparator (u), this function is designed to specifically weight house number values. The string is analyzed and the weight adjusted based on statistical data.

us 

Simplified String Comparator 

A custom string comparator that compares two strings taking into account such uncertainty factors as string length, transpositions, key punch errors, and visual memory errors. Unlike the generic string comparator (“u”), this function handles diacritical marks. This function also improves processing speed.

usu 

Language-specific String Comparator 

A custom string comparator similar to the “us” comparator with the exception that it is based in Unicode to support multiple languages and alphabets. This comparator takes one parameter indicating the language to use.

Exact char-by-char Comparator 

Compares string fields character by character. Each character must match in order for an agreement weight to be assigned.

Generic Number Comparator 

Compares numeric fields using a relative distance value to determine the match weight. As the difference between the two fields increases, the match weight decreases. Once the difference is beyond the relative distance, a disagreement weight is assigned. This comparator takes two parameters; the first indicates whether to use a relative distance or direct string comparison, and the second indicates the relative distance to use.

nI 

Integer Comparator 

Compares integer fields using a relative distance comparison. This comparison function is based on the generic number comparator (n), and accepts the same parameters.

nR 

Real Number Comparator 

Compares fields containing real numbers using a relative distance comparison. This comparison function is based on the generic number comparator (n), and accepts the same parameters.

nS 

Alphanumeric Comparator 

Compares social security numbers or other unique identifiers, taking into account any of these parameters:

  • Field length

  • Character types

  • Invalid values

dY 

Date Comparator - Year only 

Compares year values using relative distance values prior to and following the given year to determine the match weight. As the difference between the two fields increases, the match weight decreases. Once the difference is beyond the relative distance, a disagreement weight is assigned. The date comparison functions handle Gregorian years. This comparator takes up to three parameters; the first indicates whether to use a relative distance or direct string comparison, and the second and third indicate the relative distance before and after.

dM 

Date Comparator - Month-Year 

Compares the month and year using a relative distance as described above for the year comparison function (dY).

dD 

Date Comparator - Day-Month-Year 

Compares the day, month, and year using a relative distance as described above for the year comparison function (dY).

dH 

Date Comparator - Hour-Day-Month-Year 

Compares the hour, day, month, and year using a relative distance as described above for the year comparison function (dY).

dm 

Date Comparator - Min-Hour-Day- Month-Year 

Compares the minute, hour, day, month, and year using a relative distance as described above for the year comparison function (dY).

ds 

Date Comparator - Sec-Min-Hour-Day- Month-Year 

Compares the second, minute, hour, day, month, and year using a relative distance as described above for the year comparison function (dY).

Prorated Comparator 

Prorates the disagreement weight for a date or numeric field based on values you specify. Differences greater than the amount you specify receive the full disagreement weight. This comparator takes three parameters indicating the relative distance and the agreement and disagreement ranges.

The Match Constants File

The match constants file, matchConstants.cfg, defines certain configurable constants used by the match engine. This file includes four parameters, but currently only the first parameter, nFields, is used. This parameter defines the maximum number of fields being used for matching. This must be equal to or greater than the number of fields defined in the match-columns element of the Match Field file. The match constants file defines the following constants for the match engine.