- Kieran Healy, Using Metadata to Find Paul Revere
Kieran Healy’s article on finding Paul Revere with metadata was very interesting. Aside from the clever point of view of a Royal Security Administration analyst, Healy had some very interesting points to make about metadata. Basically, using only information tracing individual membership to multiple “terrorist” groups, Healy could make some really interesting and useful insights deeper into the data. Initially, the author is able to convert the table from People vs. Groups to a People vs. People table – Healy does this by multiplying the matrix (table 1) by its flipped self (transpose of table 1). This simple equation allows the author to quickly manipulate the data and to start drawing relations between different people that it might otherwise have taken quite a bit of time to discover manually. Regardless, Healy is left with links between people as they are members of the same rebel group. This works in the same way but with the multiplying matrices equation flipped. In this particular case, the author was left with a table of Groups vs. Groups, elucidating how many members each set of groups shared in common. In either case, some quick and easy data visualizations make it obvious quite quickly which groups and/or people were at the heart of the rebel colonial cause.
I believe this case study has some interesting applications in situations where your metadata collection is limited for some reason. In this case, Healy had only the information on group membership to work with, and was able to tease out some very useful relationships (which were there all along but one would probably not have picked up on without the manipulation and data visualizations). For instance, in archaeology our data sets are often limited to information like what types of objects we find and where we find them. Since we study the distant past, it is very unusual to have more information, for instance the name of the artisan who made the item, or who it belonged to. However, Healy’s methods seem to have good applicability in these cases. If for instance we could put together a spreadsheet of pottery types vs. their find locations across a large region, nation -state, or even area like the Eastern Mediterranean, then perhaps we could begin to tease out some central nodes in the data. These nodes may then correspond with production centers, and could help us to understand trade or redistribution patterns in pottery.