Finding and resolving matching entities

You can resolve the potential matches by inspecting the results and deciding what action to take. When you add data to a chart, it is important to try to ensure that multiple items do not represent the same real-world object, event, or relationship.

Use the Find Matching Entities feature to compare chart items, identify any entities that might represent the same real-world object, and resolve the potential matches. In addition to matching based on entity labels, you can also look for potential matching entities based on their semantic types, and the meaningful content. When you identify potential matches, you might want to select the matched entities for closer inspection, link the entities for inspection later, or merge the entities.

The following examples demonstrate some types of duplicate data that the Find Matching Entities feature might reveal and describe how you might choose to resolve it:
  • Several entities that have similar labels, such as "Sam STEELE", "SAMSTEELE", and "SAM steele" might exist on your chart. If you have further evidence to support the supposition that these entities all represent the same person, you can merge the entities. All of the information in the entities' properties and all the links to these entities are then stored in a single entity.
  • When you add information to a chart from different sources, data that is potentially the same might be formatted differently. For example, the entities 01234 56789 and 0123456789 might represent the same telephone number or bank account.
  • Many Address entities might have the same post code but have slightly different addresses. More investigation is required to determine whether they are duplicate locations. You can add Match links between the entities to record possible duplication and investigate further.
  • Two Person entities on your chart might have different labels, but the same date of birth and national identifier. This information might indicate that the two names are aliases for the same person. You might choose to link them and investigate further or, if you already have information that indicates that are they same person, you can merge the entities.
  1. Click the Analyze tab, and click Find Matching Entities on the ribbon menu.
    The Find Matching Entities pane opens.
  2. Select the type of matching you want from the list.
    Label Matching A specialized form of entity matching that compares entity identities and labels. No two entities can have the same identity but identities can differ by letter case or by their digits only. For example, you can search for identities that have matching letters, excluding letter case and digits.
    Smart Matching The semantic type that is assigned to the entity is used as the basis for the match. For example, the entity types Doctor, Male, and Anonymous might all be assigned the Person semantic type. Common properties of people such as national identifier and date of birth are then examined to find potential matches. Smart matching considers a broad range of criteria, including attribute instances and database properties that are assigned the same semantic types. You can specify a threshold to indicate how strong the match must be and you can choose which entities to include in the search.
    Smart Matching Against Selection Similar to Smart Matching, however chart items are matched only against the current selection on the chart surface. For example, if the entity Danny KENT is the only entity that is selected on the chart, matches are made only against Danny KENT.
    Previously Linked Matches Search your chart to find matches that you previously found and preserved on the chart by adding a Match link between them. You can find matching entities, ignoring any previously linked entities, as many times as you like. When everything of interest is linked, use Previously Linked Matches to retrieve all the resulting links from your previous searches.
  3. To set up the search criteria for the type of entity matching that you select, click Setup.
  4. Click Find.
    Any matches are displayed in a tree view in the Matched Sets area of the Find Matching Entities pane.
  5. Use the tools in the Matched Sets area to resolve the matches.
    Note: You can run Find Matching Entities as many times as you like with different options and different setups.