Matching is a helpful tool to connect your data with other providers. Metagrid provides an assistant to faster match your data. You can create a matching algorithm and click through the proposals.
Navigate to “Match” in the top right corner and click “New” to create your first matching algorithm. First, give your matching an expressive name.
Now you can start to add an “option” to your algorithm. Each option adds a parameter to compare your data with other data in Metagrid.
During the matching, Metagrid will go through all your data and compare each person using the given options against all the other persons. Metagrid uses a more-matches-is-better approach and tries to deliver results rather than returning blank.
You can add options for the names, birth, and death dates. Each option has an operator. The “must match” operator excludes results not fulfilling the option. The “should match” operator will score these results lower. If there are several results, a lower score can result in a dropout.
For the first and last name, you can specify fuzziness. Metagrid uses the Levenshtein algorithm to calculate the fuzziness. With a fuzziness of 1, one character of a word can differ - f.e. Miller and Millar would match.
For dates, you can specify a precision and a tolerance. With both arguments, you can define a range in which the date would match - e.g. with precision “year” and tolerance of 1 1955-10-05 and 1954-01-01 would match.
Test your algorithm when you are ready. The test matches 200 randomly selected persons of your provider and displays the result. If everything looks good, run the algorithm. Running the algorithm takes some time, so be patient. You can check the state of the matching on the Match overview site.
After the matching has finished, you can work through the proposals. To accept a suggestion, click on the switch button in the header and save. Metagrid will link the selected proposals together and store the refused matches.
When clicking on the gear icon, you can view the options chosen for that specific matching.