For the past year or so I’ve been using Palladio for some data analysis alongside R and Gephi. The developers call it a “web-based platform for the visualisation of complex, multi dimensional data”. It essentially takes tables of data, visualises them, and allows you to link them to other tables. This last bit is the most exciting: you can have one table with a list of elements (say, people, and the links between them) and link it to another table with other information (dates of births, say, or location information). Then you can make visual mash-ups of both tables, or even add more! It even recognises types of data and automatically displays them correctly!
Here’s my use: I have lists of connections in one table, and the data they occur. Many of them are repeated. My first table looks like this:
1st Location: 2nd Location: Date:
Venice Augsburg 01/05/1647
And another 1,200 rows, with many repeated sets of connections.
Palladio allows me to upload a second table and link it to the first. So I have a second table, which contains latitude and longitude coordinates for each place. When I link the two, the software ‘encodes’ each instance of a place with the correct coordinates. It saves on massive amounts of duplicated information, for one thing. Imagine if my first table had a million rows?
This is basically the theory of a relational database, which have been around for years, but with Palladio you can instantly visualise everything, which is pretty groundbreaking. It’s all very well having linked tables in Microsoft Access or on an SQL server but particularly to a digital humanist, the data isn’t really worth much until you can figure out easy ways to interpret and analyse it. Data for us is the beginning, not the end.
Palladio is incredibly simple to use – to the point where I presumed I was missing a step when all I needed to do to get started was upload a couple of .csv files to a web app. Within a few minutes of opening the software I could create a visualisation like this:
The really interesting aspect for me is this timeline at the bottom, which allows me to do something I thought pretty much impossible without writing bespoke code: filter network connections by date (Gephi collapses all repeated connections into one), and it provides a really nice interface through which to view it. It’s also quite beautiful: all monochrome and minimal and transparent. That’s important too!
However, I have a couple of suggestions I’d like to make for future revisions. Some of which are better thought out than others, and I’d like to make it clear that I’m amazingly appreciative of the hard work the developers have done on this so far!
Allow facets to be applied to layers:
I’d love to be able to add one layer with my data filtered in one way (say, all the connections from a date range) and overlay it with another to compare. For example, I would love to be able to compare how my network looked like in 1649 compared to 1645! There are difficulties in this: there would need to be some way to distinguish nodes when they overlap, which brings me to my second suggestion:
Allow users to change the alpha transparency of facets:
I thought maybe if you could set both the colour and alpha transparency of individual facets it would get around the problem of indistinguishable overlapping nodes. At the moment you can change the colour, but it doesn’t have any value over aesthetic. If you could set the transparency and colour of each of your facets, you could have multiple displayed at once, seeing the smaller ones underneath. It would be a great way of comparing several versions of a network or other dataset at once. Also, mapping transparency is really helpful with huge datasets – allowing patterns to be seen through complexity that is impossible otherwise.
Allow aesthetics to be mapped to variables:
Sort of reaching dreamland with this one – if Palladio was a Kickstarter project this would definitely be a ‘stretch goal’, but might as well shoot for the stars I guess… It would be amazing to have some of the functionality of R in this regard. R allows you to map a large number of aesthetics to pretty much any variable. This would be particularly powerful with relational databases. I can imagine having a table with connections, say of people, and a separate table with some numerical property – their net worth, perhaps. Being able to set the colour shade or transparency to this property would be a great way of visualising extra information.
The ability to publish/embed Palladio files:
Not going to say too much about this because I think it gets asked a lot and it sounds like the team are working on it. But we historians are a sharing bunch! It would be amazing to share interactive charts with other historians and the public. Palladio could potentially even be a useful tool for journalists, sort of a D3.js with an easier learning curve.
Make it work better on a small screen:
Palladio is a little unwieldy to use on a laptop. For example, on a 13 inch laptop, even at full screen, I can’t click ‘apply’ to make changes to a layer: it’s hidden behind the bottom toolbar. More of a bug fix than a suggestion. Perhaps having the option to move elements of the page around would help with this. Having the timeline on the left as a vertical bar would free up more useful screen real-estate.
That’s it! If you haven’t given Palladio a go, I highly recommend it if you have some structured data that you’d like to visualise. At some point I’ll put up a post actually talking through some of the network properties I’ve found through using the software.