Palladio is a web-based software created at Stanford for use in creating simple web-based network graphs. It is great for use for less complex graphs since there is no software to download and putting in both the nodes and the links between them is as easy as copying and pasting information into a box. When the graph is created, filtering it down to only the nodes with certain attributes is a cinch as well.
However, if you are looking for a graph you can host on your own website, or if you want to customize your graph beyond just making one set of nodes larger than another based on a few options of measurement, this program is too unsophisticated for your needs. You can take a screenshot, but you can't embed your graph and host it somewhere else. While you can make the nodes a different size based on the amount of connections they have or amount of nodes they are connected to, you can't make nodes of different categories distinct in their own ways, so if you're trying to display multiple kinds of information about what you are creating, Gephi or NodeXL might be more fitting for you.
When you're making a network analysis graph, the initial data that you work from will either be structured or unstructured. If it's structured, that means that it's already some kind of table or chart where who is connected to who is clearly listed, as well as any other pertinent data about that connection. If you were trying to do a network analysis of campaign donations and the data you got was formatted something like the below table, it would be structured.
Donor | Candidate | Amount |
Nerese Campbell | Clarence Royce | 300 |
Stringer Bell | Clay Davis | 5000 |
Andy Krawczyk | Clay Davis | 5000 |
Andy Krawczyk | Clarence Royce | 3000 |
Because you received your information on campaign donations in a form where it's clearly marked who the money is from (one node in a connection), the donation (the link between two nodes) and the person it is going to (the other node in the connection), that is structured data. It is easier to start from structured data since you can skip several steps, but this isn't always possible. Maybe instead of having the data this explicitly laid out for you, you only have pictures from fundraisers that were published in the paper, or politician's memoirs or a series of articles containing donor and candidate information. That data would be unstructured. In those cases, you'd have to organize it yourself to conclude what entities are connected, what kind of connections are present, and what additional attributes you want to be available for each of these people and connections as you make your graph.
With this exercise, you'll take the text of A Midsummer Night's Dream and do a network analysis of what characters share the stage in the first act and get it ready to put into the network graphing platform Palladio. This is more like semi-structured data since the play clearly breaks out what characters are on stage and when scenes end so it's a little easier to start with when converting a work into structured data.
I took the text from Project Gutenberg, a nonprofit that puts classic works up online. You can download the text file if you want, but I find the link below most helpful, since it provides a list of players and breaks the speech up in a clear fashion.
A network graph consists of nodes and the links between them. When you have unstructured text, you'll need to decide what you'll count as nodes and links within it, and this is called a making a coding scheme. You'll make this coding scheme by skimming the work to see what kind of information it has about your research question and how you'll want to categorize that information. Your graph can be supplemented by attributes for both the nodes and links. Attributes are categories or other information about the nodes and links between them, and so you'll want to decide if there are other attributes about those nodes and links that you'll want to record as you're making your coding scheme. After you have the rules set for yourself, you go through the work recording what's in it using your coding scheme. Basically you decide what to count and when, and then you use those rules to count the connections in the work you want to analyze. But first you need to create the coding scheme.
Take a look at the play and ask yourself:
How you'll answer these questions will depend on what you are hoping to find out from your network graph. If you were interested in how much characters were being talked about in a play behind their backs, you might only record mentions of one character who is not onstage by another character as a link, and you'd count it based on the number of mentions. If your research question you wanted to answer is about which characters directly speak to which other characters, then you'd count each line as a link between those two characters, and disregard any soliloquies or lines where it's not clear who a character is talking to. Your research question also determines what attribute information you want to record about those nodes and links.
In our scenario, you're just going to be trying to answer a research question about characters in different groups appearing together, and so only need to record a link if the characters are on stage at the same time.
Basically what you'll be aiming to do with your coding scheme is gathering enough information about your nodes and links that you can filter your data in ways that will help you answer your research question but not too much that you spend a lot of time gathering information that you don't need.
In this case, we just want to look at the appearances of characters together, so we won't be entering information on people or other items that are mentioned but never appear, nor the places mentioned in connection with these people.
Whenever you can, you'll want to have an official or outside source of who the nodes should be in your graph. When you decide the scope that you are using, you'll want to have a reason to argue for why you made it as narrow or broad as you did. Handily, in this case, you do have a list of dramatis personae at the beginning of the play.
The characters listed in bold seem to be the most important to the story, but you should look at them to see if there are any exceptions you wouldn't want to include. For instance, in A Midsummer Night's Dream, there is a play within the play performed by one group of characters for another called Pyramus and Thisbe. In this list of characters, the ones in Pyramus and Thisbe are included so you'd need to decide if you want to have those characters as separate nodes on the graph from the Clowns (or Mechanicals) playing them.
In this case I'll make the decision of no, and if we get to a scene where Bottom is playing Pyramus and Flute is playing Thisbe, I'll mark it down as a connection between Bottom and Flute, rather than Pyramus and Thisbe. As for the fairies and attendants who go unnamed in the cast list, I'll record them as well since my research question is interested in how many times characters are connected to other characters as well as how many characters they are linked to.
Now that you've decided who you'll be collecting information on, you can start your node sheet
Node |
Theseus |
Egeus |
Lysander |
Demetrius |
Philostrate |
Quince |
Snug |
Bottom |
Flute |
Snout |
Starveling |
Hippolyta |
Hermia |
Helena |
Oberon |
Titania |
Puck |
Peasblossom |
Cobweb |
Moth |
Mustardseed |
Fairy |
Attendant |
At this stage you'll also want to decide what attributes of these nodes you'll want to record. Here are the two I've picked:
If there is additional information that you want to record on any of these nodes, please feel free to add more categories, but my Nodes sheet now looks like this.
Node | Group | Gender |
Theseus | Royals | Man |
Egeus | Royals | Man |
Lysander | Royals | Man |
Demetrius | Royals | Man |
Philostrate | Royals | Man |
Quince | Mechanicals | Man |
Snug | Mechanicals | Man |
Bottom | Mechanicals | Man |
Flute | Mechanicals | Man |
Snout | Mechanicals | Man |
Starveling | Mechanicals | Man |
Hippolyta | Royals | Woman |
Hermia | Royals | Woman |
Helena | Royals | Woman |
Oberon | Fairies | Man |
Titania | Fairies | Woman |
Puck | Fairies | Unspecified |
Peasblossom | Fairies | Man |
Cobweb | Fairies | Man |
Moth | Fairies | Unspecified |
Mustardseed | Fairies | Man |
Now that you've figured out the nodes that will be on your graph, you can move on to how you'll be recording their links with each other. You may find as you start coding that you'll want additional information on these characters recorded, just be open to amending your sheet if you need to, it's much easier to revise this sheet than the Links sheet.
Developing a Coding Scheme - Links & Link Attributes
A network graph isn't just about the entities involved, but the links that connect them. You'll be deciding here what links will be counted and what information you'll be recording on them. As above the first thing to consider is what you're trying to visualize with your graph, which is which groups of characters appear in scenes together. While another kind of network graph might show that Egeus is Hermia's father and Oberon is Titania's husband, this is one that is focused only on who shares stage time with who, and how often.
Stage Mate 1 | Stage Mate 2 |
Titania | Oberon |
Oberon | Titania |
We decided to only record character as linked when they are the ones onstage together, not when they are mentioned by another character when offstage. This means on the Links sheet, you don't need an additional column describing the nature of the link. However, even though you'll only be coding the first act in this exercise, if you were doing the whole play, you'd probably want to know if the density of the network changes from one act to another, so you'll want to add additional attributes for the part of the play the links occur in.
You now have your criteria set up to start coding the unstructured data of the first act of A Midsummer Night's Dream.
You'll be taking each grouping of people onstage at a time and adding that data to your Links sheet. When you do this on an actual project you may find that starting to code your data will teach you that something about the methodology you began with didn't take into account something important about your data. Maybe you'll find that you've left out a kind of connection that you think is important, like if you got a scene or two into the play and realized that there actually were a lot of mentions of characters when they were offstage so it shouldn't be something you ignore. It is okay to let your project evolve this way, but if you'd decided that mentions of a character who is offstage is something you want to record, you'd need to add a column that let you specify if the link was from a mention or from being onstage with a character, then go back and record that information from the beginning of the play.
Let's start with this passage, the opening of the play.
SCENE I. Athens. A room in the Palace of THESEUS
[Enter THESEUS, HIPPOLYTA, PHILOSTRATE, and Attendants.]
THESEUS
Now, fair Hippolyta, our nuptial hour
Draws on apace; four happy days bring in
Another moon; but, oh, methinks, how slow
This old moon wanes! She lingers my desires,
Like to a step-dame or a dowager,
Long withering out a young man's revenue.
HIPPOLYTA
Four days will quickly steep themselves in nights;
Four nights will quickly dream away the time;
And then the moon, like to a silver bow
New bent in heaven, shall behold the night
Of our solemnities.
THESEUS
Go, Philostrate,
Stir up the Athenian youth to merriments;
Awake the pert and nimble spirit of mirth;
Turn melancholy forth to funerals—
The pale companion is not for our pomp.—
[Exit PHILOSTRATE.]
Hippolyta, I woo'd thee with my sword,
And won thy love doing thee injuries;
But I will wed thee in another key,
With pomp, with triumph, and with revelling.
[Enter EGEUS, HERMIA, LYSANDER, and DEMETRIUS.]
When you're done with the other acts, you are ready to move into the next module and graph the network of these characters in this play. If you get lost and don't know how I would have classified a scene, go to the link below.
In the first tab of this tutorial, you took Act I (or the whole play) of A Midsummer Night's Dream and recorded the data it contained about which characters were onstage with each other at the same time into a structured spreadsheet you created. With this data you'll create a network graph in Palladio of how often those characters are onstage with each other. You'll also be able to sort those relations by where they are chronologically in the play, what group each character is in, and what gender they are since that's information that you've recorded.
If you didn't create the data in the last module, you'll find it here.
Palladio makes it very easy to import your data. You can either upload csv sheets or simply cut and paste them into the interface. This is the first step to Palladio making a network graph with that data.
Now that you've entered in all your information, you can see what the connections between these different characters look like in visual form by asking Palladio to create a graph.
When filling out the Nodes and Links sheet, you put in additional attributes. For the links, this is what act and scene this connection occurred in. For the nodes, this was information about which group the character fell into and what gender they were. You can use the facets in order to filter which nodes and connections appear in the graph, which depending on your research question can be quite useful.
Unfortunately, Palladio doesn't offer an option for you to embed the graph, but there are a couple of different ways that you can save it. For instance if you want your graph as a json file that you can then feed into a graphing or other visualization program, you can click on the download icon at the top of the chart. Note that this will download the entire information for the links and nodes sheets cross-referenced, not just whatever information that you have filtered and configured as your graph.
If you want an image of any of the graphs that you've made, just click on the Download icon at the bottom of the Settings menu. This will download an svg image to your computer. Then you can paste it within your paper.
Though Palladio isn't the most customizable or detailed way of doing network analysis, it is a handy tool if you're just curious about what connections are in your data before embarking on a more technical way to graph your data. It also provides illustrations with a minimum of fuss..