Computational journalism in practice

A crowd-tagged map from just one of the breakout groups from the first round of community workshops. The green dots represent 'strong places' and the red dots represent 'weak places.'

Recently I wrote a piece on the Yellow Springs/Miami Township visioning process for the paper I write for, the Yellow Springs News. I was working with a large amount of pre-sorted ideas (narrative phrases generated by community members and recorded verbatim by session leaders) to summarize the draft findings from the first community workshops. The article was in anticipation of the kick-off to the second stage of the process, another public workshop designed to help community members work some of these raw ideas into goals.

But narrative phrases that have been pre-sorted into categories (with one idea sometimes being filed under multiple categories) are fairly difficult to assess. From a journalistic perspective, at first glance the ideas seemed oversimplified. Looking closer, the ideas took on a disembodied quality—especially when considering the ‘weak place/strong place’ results of the community mapping project, one of two exercises conducted at the first-round workshops. Since my job as a writer is to make sense of things for my reader in the clearest way possible (as my editors will remind me!), I had to find a way to get deeper into the community generated data—in a nimble kind of way. I was on deadline.

Enter Many Eyes, a set of text mining tools anyone can use to visualize data—with the hopes of revealing unexpected patterns and insights that would be difficult (impossible!) to wrap one’s mind around without computational assistance. Since data visualization is key to my interview work here (I consider Wordle to be a primary research tool/method in my community conversation section), I’ve been looking for a way to use Many Eyes. But because all data sets need to be open-sourced for the greater Many Eyes community, it is not quite appropriate for my oral history/ethnography interviews. So I took the plunge and uploaded two data sets provided to the community by ACP Visioning & Planning, the consultants hired by Village Council to lead the visioning process. (That is, I copied and pasted text into a window.)

Because I participated in the first round of workshops, I had a feeling that some of the mapped responses would be a bit unwieldy. For starters, the dots were very large compared to the scale of the map. And community member’s thinking behind how to place the dots seemed to vary. For instance, after the mapping was finished, one community member said he placed a weak dot on Antioch College because it needs development. Other community members placed green dots on Antioch College, because it was an asset or a staple of the community. Likewise, while many tagged the western gateway of the village as a weak place, a few tagged it green because of its potential.

Using the word tree tool, I was able to see how people referenced these places during their comments that were recorded by the session leader. While the map helps us get a visual handle on things, the tree map allows us to consider what that visual means, by considering those key place words in context. In the weak data set, sorted by ACP, many comments were recorded about the western gateway to the community. You can enter your own search terms, and navigate around from there. Each time you click on a new term, the word tree is remapped. Try ‘western gateway,’ ‘Mcgregor,’ or ‘Vernay’ to get a sense of how to the word map allows you to move through the data.

Weak Set

The strong set, as sorted by ACP, included many responses about the Glen, Antioch College, and downtown. Grab a search term, and see how it moves for you. As a first time attempt with the word tree tool, with copied and pasted data that was not collected for the purpose, I was pretty impressed. My needs as a journalist were far better served through this method than flipping through flat pages of text and power-point files. Since community generated data is non-linear, it makes sense that my research process ought not be linear either.

Grab some search terms and see how it moves for you. And if that gets you thinking, check out these data sets.

Strong Set