Composite Matching Process

The Composite matching process includes matching incoming records in a composite entity. The composite entity refers to an index entity, where two or more related repositories are combined together.

  • Person > {Relationship: ResidenceAddress} > Address

In this case, each index document refers to a record in the Person repository related by ResidenceAddress relationship to a record in the Address repository.

When a new Person record related to the Address record is added in the application, the MatchRecord activity first identifies whether an eligible index entity exists in the Netrics index that satisfies the following conditions:

Whether an index entity includes Person and Address repositories related by the ResidenceAddress

  • If yes, this is the most eligible entity. Refer to step 3.
  • If no, refer to the next step.

    Whether an index entity includes the root of the record bundle, that is the Person repository

  • If yes, this is the most eligible entity. Refer to step 3.
  • If no, refer to the next step.

    If the Netrics index does not include any eligible index entity to query, the search cannot be performed. The MatchRecord activity flags the new record (bundle) as an accepted record without detecting any duplicates.

    If an eligible index entity is found, verify the attributes that forms the index entity.

    For each of the attributes, locate the appropriate record values from the record bundle.

    Flatten the record bundle to form the appropriate structure. For example,

    Person.FirstName Person.LastName Address.City Address.Zip
    "John" "Doe" "Palo Alto" 94040

    Create a fuzzy (non-deterministic) query that includes the attribute-value combination, any specified weights, and minimum matching threshold value.

    Go to the Index search results for the specified fuzzy query:

  • If no results are found, then the record bundle does not contain any duplicates.
  • If one result is found, then the record bundle contains one duplicate bundle.
  • If more than one results are found, then the record bundle contains multiple duplicates.

    Save the search results in database and cache.

For more information on the Data Quality process after the Matching process, refer to the Process Definition section in TIBCO MDM Customization Guide.