Warning: this post is NOT intended as being a tutorial on how to start using Gephi. I’m just giving you my takes on some of the updates of version 0.9. If you are looking for a tutorial, go there (for a very complete tutorial) then then here for some tools’n’tips.
In academia, journalism and on the web in general, we see more and more of those colored circles with lines joining them. Some people call them networks. Some even believe that these networks —when modeled, decoded and analyzed correctly—, contain the Holy grail for scientists and journalists: explanatory power. To play with data and networks I usually use Gephi, which is a tool that has the benefit of being relatively easy to handle and of being OpenSource. People at Gephi describe their software as:
An interactive visualization and exploration platform for all kinds of networks and complex systems, dynamic and hierarchical graphs. It runs on Windows, Mac OS X and Linux. Gephi is open-source and free. [More at Gephi.org].
So Why should we talk about Gephi today? Well because Gephi release its version 0.9. It should be one of the last versions of the before the version 1.0… I guess? Until a few days ago, Gephi ran its version 0.8.2. It might not seem like a big change BUT IT IS. The complete list of changes can be found here.
what exactly has changed in this version?!
Well I’m no IT guy so I can’t tell that the compatibility of the debug 010101 java extend processus has now been pushed full speed by bypassing hardcoded modules or anything like that, but I can tell you that it just runs better. Everything is smoother, faster and more user-friendly. I used to loose myself in the software for hours just to find a way to do this one little thing to my data, to change the graphs’ appearances, etc. Now any scientist or anyone who is interested in producing stories with networks can do so without prior knowledge to network analysis. Let’s develop a little what I think of the new version.
What I appreciate about it
- Fluidity. As stated above, everything is smoother and more fluid. The spatializations —ways of moving the nodes automatically using pre-made algorithms— work faster. The preview mode works better too. Playing around with networks really became playing around in this version.
- Data laboratory. The data lab, (where you work with your data, connect nodes with edges, etc.) has been improved greatly. It is now easier to change the names of the nodes and so on. You are now able to set “times” (instead of “timestamps”) which means that you can easily create a dynamic graph that evolves over time, such as the evolution of posts related to a topic on social media (which I still have to try on this version). Another example: by right clicking on a node in the data lab or in the other mode, you can directly access de the nodes’ contents on a website (if the node is an URL such as a website or a post on social media). You can also tag nodes.
- Colors, colors, colors ! It might look like a detail, but it is one of the most important things to me in this version of Gephi. Now there are schemes of colors that are pre-coded just for you. The previous version required to find which colors would fit together. That meant for people that didn’t have time to loose or knowledge in colors’ combination that your graphs were ugly as shit. Here is a before and after of a graph. (Labels have been deleted for anonymity but It shows people in a municipality of Brussels talking about the Transatlantic Trade and Investment Partnership – TTIP):
- Again, this might sound like a detail but it’s not. Network visualisation is all about making patterns, links and clusters stand out to understand relationships between entities. Colors, sizes and positions of nodes and links do matter. If you are looking for inspiration, go check out what this guy does, it is great. He uses Gephi amongst other softwares and produces great visualisations and analyzes. By the way, it is by looking at some of his graphs that I discovered the power of only using one or maximum 2 colors and of “color inversion”… That’s how I call it anyways. Instead coloring the big nodes in a darker color (which is the usual for some reason) you invert it, making the big nodes white. This is where I discovered the idea. Anyways, the way you color nodes always depends on various factors; the most important ones to me being the number of nodes in your network as well as what you are trying to show/say with your graph.
What still needs some work
There are still some things that do not work properly or could be bettered according to me. Naturally, I’m not saying this just to be mean or because there is anything I could do about it. However, knowing these issues will save you some time and nervous breakdowns.
- There is still no back button or cmd/ctrl+z. This means that when you do something, you can’t go back. Did you really liked the way your network looked before you changed the looks of it or before you ran this spatialization? Too bad, it’s never coming back. The only way to prevent yourself from having a lot of regrets is to save your graph before you run an important spatialization or do something big in your data lab. With time, you’ll get used to it.
- There are still some glitches and twitches (at least on my computer). Again, if you save regularly and re-start the software you shouldn’t loose too much time. It’s a good thing to do with any software anyways.
- The Workspaces mystery. The workspaces (the ability to open multiple tabs in the software) was moved and has changed somehow. I’m still trying to work it out. So far, I can open a new “tab” in the workspaces menu but when I try to open a new graph, the old one disappears… The solution I found is to export CSV tables of nodes and edges of my previous graphs, and import them in a new tab. Then it works for me. But it takes so much time that it might not be worth it to you… You can just work your graphs separately.
- Plug-ins!!! So far plug-ins don’t seem to work. I do think that this is just a compatibility issue and that it will be fixed when people just “update” their plug-ins. I sure hope so, as plug-ins are essential to Gephi (to export visualisations to the web, use new spatializations, and altogether customize the software).
To wrap things up, I’d say that I think user-friendlyness is something essential for Gephi. This update makes Gephi accessible to a larger audience. I have no doubts it will attract new users, people will chat, build up the community, find new ideas and improve DataViz altogether. For this, and all the rest, I give a big thank you to the people at Gephi !
Why ? Because I can. Until next time, stay golden.