An early IPTN project involved scoping how the term 'peacetech' had been used over the last 30 years. We wondered what technologies have been used for peacebuilding and how this varied over time; and how - if at all - technological developments and peace processes drive one another.
One way to do this was to use data visualisation programmes to look for patterns in the use of terms and technologies across multipe articles. Madlen Williams, SEO at the Jean Golding Institute and digital anthropologist, and Huw Day, data scientist at the Jean Golding Institute, worked together on the project.
Madlen: ‘Visualising patterns in data – here, in literature - is a completely different way to understand research for me. Along the way, I learnt that certain data sets (hierarchical, relational) limit visualisation options, while Huw’s method to search simultaneously through multiple documents for desired terms hugely cut down my workload. An excellent example of how data science can impact research across disciplines - even those without a data science background.’
Huw: ‘I have never really approached reviewing literature in a quantitative way before. To understand what visualisation tools we would need to understand the literature better required an iterative approach – trial and rethink – data collection with a tech fix.’

How they went about it
We identified key tech terms that we wanted to track the presence of in the peacetech literature, such as internet, VR, GIS. Madlen searched for the mention of these terms in papers, reports and case studies, verifying that each mention was contextually relevant. For example, if a paper said: "In this example we avoid the use of radio communication by using the internet", we would only associate the word "internet" with this particular paper unless "radio" came up later in the paper in another context.
We considered an automated search using natural language processing techniques (using something like ctrl-F to find mention of words in documents), but the nuanced problem of verifying that each term has relevance in its particular context made it to complex for this limited project.
Once all the literature had been manually searched for key terms, data visualisations were generated that described the number of papers that mentioned a particular tech term. The main mechanism for this was a stacked histogram plot, where the literature was grouped by year (in five-year length chunks) and the papers that mentioned each tech term were counted.
Those numbers were then plotted onto the vertical axis, with different colours representing different tech terms. The bump chart on the left is just one example of how you could visualise this trend over time, showing the chronological development of tech tools cited within the data.