Entity linker

The EntLinker is the primary facilitator of the linking process.

Each time a record is created or updated, the MDM operational server cross-matches that record against other records in the MDM database. Parameters for entity management and creation direct the EntLinker on how records are to be processed. (The parameters are defined through the Member Type > Entities editor in the InfoSphere® MDM Workbench MDM Configuration perspective.) In addition to the parameters that control entity manager polling and work units, the configuration dictates whether to:

• Use an input work queue (EntIque) (input queues are used if the MDM operational server is configured to create linkages and tasks)
• Allow cross-matching for member records with a status of Active, Merged, Obsolete, Deleted, or a combination of statuses
• Allow linking of records from the same source
• Allow group entities
• Create linkages and tasks
• Store identity rules for records in an entity

The EntLinker associates records into entities that are determined after a comparison of record attributes. The comparison scores, with the parameter settings mentioned earlier, and the defined thresholds result in the EntLinker:

• Auto-linking the records and assigning a common Entity ID (thus creating an entity),
• Creating tasks for the records that score between the auto-link and clerical review thresholds,
• Or doing nothing if the records score below the clerical review threshold or if the configuration does not allow for linkages or tasks.

Identity and non-identity rules also affect how the EntLinker processes records. Identity rules are pair-based rules that are set between two members. Rules between members override the linkage decisions that are made by the operational server during comparison. For example, records for two members who are twins always compare and score high. A user can determine that the two records are different individuals and can set a non-identity rule by using the linkage and relationship functions in the Inspector application. This rule now prevents the records from being linked during future comparisons. An identity rule results from manually linking two records. When an identity rule is set, the records will not unlink in future comparisons regardless of the comparison score.

Linkage information is stored in the mpi_entlink table. The member and entity rules that are established during linking are stored in the mpi_memrule and mpi_entrule tables. Entity ID assignment history is stored in the mpi_entxeia table, while a history of modifications to member linkages is stored in mpi_memxeia.