Recently I was lucky enough to be invited to present findings about street children at the University of Manchester in UK. Eleanor Harrison (GlobalGiving UK CEO and former director of a street children program in Kenya) yielded her time to me, and I gave this presentation, virtually analyzing all the data on the spot.
Not being able to prepare in advance was a blessing. It helped me emphasize that we’ve come a long way in our quest to build simple, instant, intuitive tools for analyzing narratives and their meta data. These images are screen shots of the tools I’ve described previously at djotjog.com/search and djotjog.com/report.
What can you do with noisy, minimalist, or totally unstructured narratives?
A demonstration with stories about street children
First, it provides a top level visual summary
You don’t even need to know much English. Just look at the pictures and you’ll get an overall idea. Everybody talks about street children, but kids talk about it a bit more than the rest.
Analyze the point of view: “Actor in the story” versus “Affected by events”
When people tell their own stories, they turn out more negative. But young boys are more positive and young girls are more negative.
A story’s point of view affects the depth, nuance, detail, and likely “data” that can be extracted:
Street children stories have more “I” stories that we would expect, pulling stories from our collection of 60,000+ at random, so the “I” is bigger.
A Wordtree is an unsupervised algorithm that works with any text
Djotjog.com/search generates these for ANY search result in seconds. Here are branches from hundreds of street children stories.
Zooming in you can see how words and phrases connect to form ideas: An old man frequently interacting with a young girl, with money being part of that sentence in the story.
Food and life also form two branches of the story.
We can look at how stories intersect with mentions of specific organizations.
Contrasting stories with different endings can yield more program-level design insights. (Using stories about success or failure we map outcomes)
This is the most interesting snapshot of them all. Failure stories are more complex than success stories. Wordtrees look for associated words within sentences. Success stories have branches that fan outward because the concepts are simpler, less interwoven. Failure stories are a chaotic birdsnest of interwoven social issues.
In failure stories, each storyteller tends to describe the events using similar words but in a different order so the map is simply more complex, more interesting.
As organizations, too often we focus on learning from the success stories – but it is the failure stories that offer us the keys to understanding the problem.
Compare two story collections
Side by side: Stories from Pennsylvania (N= approx 60) and East Africa about street children / run aways are remarkably similar in the issues they address.
And here is a map of about 50 blogs posts about street girls in Cairo form Nelly Ali’s blog:
You can do this yourself. Go to djotjog.com/search and type
"street children" or "street kids" or "run away" or "ran away"
in the search box, then hit “fetch stories.”
Note: you can export the narratives and meta data as a CSV file, for further exploration in a tool such as SAS, SPSS, or BigML.com.
Postscript: What MORE can you do with narratives? Try BigML.com
I imported hundreds of stories that mention ‘child abuse,’ ‘child labor,’ or ‘child protection’ into BigML.com and used their version of a text-mining function to build trees that map outcomes in stories. The types of outcomes are how people felt about the story they told, such as ‘inspired’, ‘horrible’, or ‘happy:
Horrible stories
Inspiring stories
Have mixed outcomes, and don’t mention your aunt. But more importantly, they are not from people who were “affected” by the events in the stories they share.
Important stories
Are commonly organization-centric narratives.
Happy stories
Have a very long pattern of words in them. This example begins to demonstrate the true scope and nuance of “success” that lies in narratives, and which cannot be captured with a simple “did the right people benefit” question.
One thought on “Understanding street children: What can you do with narratives?”