Anonymize identifying information without losing data

[yewborn]

I have a large dataset (70,000 rows) on property ownership, which includes personal identifying information (i.e. people's names). I want to make the data anonymous so that other users cannot identify the property owners. However, I need to preserve the internal fidelity of the names, so that users can tell if multiple properties have the same owner. To further complicate things, the data are littered with typos. So the same name may appear several times, but spelled slightly differently. Here's an example:

Property number	Owner name
1	Jeremiah Wilson
2	Emily Chang
3	Emily Chang
4	Jeremaih Wilson
5	Jeremiah W. Brown

In the above example data, I want to achieve the following:

• the names in Column 2 cannot be identified
• the names for properties 2 and 3 are identical to each other
• the names for properties 1 and 4 are similar enough to know this is a typo
• the names for property 1 and 5 are clearly different

In an ideal world, the solution would be relatively simple to implement. I am requesting these data from a government agency, and I need to provide them instructions for how to anonymize the data while preserving the internal fidelity of the names.

I apologize if I'm not using the correct terminology for aspects of this problem. Thank you for your help.