Several organizations in Kenya and Uganda (RETRAK, TYSA, Zawadi fund, have requested a detailed analysis of emergent themes about street children from our collection of 40,000 stories. Here is what I came up with:
Tool: auto-suggests related themes
First, I pulled all stories that contained any one of these words: ‘street’, ‘run away’, ‘ran away’, ‘homeless’ and ended up with 1684 matches out of ~40,000. My program revealed that these were the most common 2-word phrases among 1684 matching stories (and number of times each phrase appeared):
ran away: 99 run away: 89 street children: 54 going school: 49 go school: 33 left homeless: 29 helped children: 27 take care: 26 lost parents: 26 helped street: 26
Auto-suggesting emerging phrases important to NGOs
Using an alternative phrase extraction method reveals the most common 2-word phrases that match keywords from one of the 10,000 globalgiving projects. (So these phrases are likely to be interesting to organizations, and you can see “street children” and “street kids” are #1 and #2 most frequent among our stories, so these are clearly relevant stories ) street children street kids provide education disabled children child abuse save lives youth sports child mother saving lives child protection hope life young girls child care orphans vulnerable based organization. school fees primary school Side note: That auto-suggestion tool I just demonstrated took 10 seconds to process 1684 stories and can be used to expand or refine the group of stories. This is really important when topics are fuzzy, like ‘human rights.’ For this analysis, I was satisfied that this set of stories was relevant and cohesive. This is an example of algorithm-aided analysis. I guess at phrases, the computer provides feedback, and I adjust the search criteria until stories have been correctly categorized as relevant or not relevant. Social scientists should always used computers to get a ‘reality check’.
Wordle provides basic summary of runaway story topics
Gephi provides more with word-connection maps
Gephi is another free tool that can map word associations. Below is a summary of all words that appeared at least 10 times in these 1684 stories, and which words either preceded or followed each word. It’s more readable than a wordle. Within this overall map are some interesting clusters. First, parents are very important, and appear to be causal factor, as parents is connected to ‘died’, ‘lost’,’passed’ and ‘died’ is connected to ‘mother’ or ‘father’. This is pretty compelling evidence that death of a parent is a specific risk factor for kids running away from home. Child Abuse is also present, but I’ll get back to that in a moment. And what actions do people take in these stories to help street children? And just so you see that the direct approach of tagging the phrase ‘street’ (as in ‘street children’) would not reveal as much in adjacent words, here is a focus on that part of the map. The only interesting word connected with street children in sentences from stories is orphan.
Gephi word-trees version 2.0
Using a different algorithm that “walks” through stories and plucks out emergent subthemes, you can detect different patterns. For example, stories of run away girls and boys branch out in very different ways. Stories that mention drugs also mention HIV, and are associated with the work of on organization more than the rest. Education is a major theme of is an ’embedded’ part of all the other themes, and appears in the center. This pulls words from 789 stories that contain one of these phrases: “street kids” “street children” “run away” “ran away”:
Does child abuse contribute to street children?
Using some of the other questions authors answered about their stories, here is a wordle of all the text from stories by 30 people who were affected by the events in their stories, and told “failure” stories (the wrong people or nobody benefited from the intervention): It is not clear that child abuse is a major driving factor. However, it is ONE of the factors as seen in the gephi word-connection plot.
Do success stories differ from failure stories in topics?
Here are words from 662 success stories about street children: And 90 failure stories: I also mapped 100 stories that were in between success and filaure. The only strong pattern is that the frequency of the use of the phrase “street children” goes down in failure stories (and remains high in the in-between stories). Also, government is mentioned as the principal actor in failure stories much more often than in success stories.
Who is addressing the needs of street children?
This image is a montage of story data parsed by type of organization mentioned in the story using SenseMaker(R) by www.Cognitive-Edge.com. Each column shows one dot per story, on a scale between success and failure (which is a composite of 3 survey questions). The color of the dot relates to the storyteller’s perspective. Most stories are from observers or those affected by the events they talk about. As the SUCCESS-FAILURE scales illustrate, few failure stories are attributed to international or local NGOs, and international NGOs are not working with street children (just 20 of 852 stories). Local NGOs are more prevalent, but most of the stories are about individuals doing something about the problem, or nobody was named. When government is mentioned, more of these stories appear to be failure stories.
Reference data: Runaway Lives (testimonials collected at Penn State University)
I mined all the narratives from this website http://www2.lv.psu.edu/jkl1/runawaylives/readstories.html and placed them into a Gephi word-phrase network map: ALL of these 60 stories are first-hand accounts from run aways in USA. Some of the same themes emerge here, such as drugs, abuse, parents, and sex — but unlike those stories from Kenya and Uganda, friends and brothers / sisters appear more prominently. What role do siblings play in lives of African street children? Mom is an important actor in American run away stories. Dad / father / papa is absent from both versions of the street-child narrative. The only time dad gets mentioned is when he dies. I hope this will spark some debate about how we extend an analysis like this. For example, I’d like to use network word-plots to contrast success and failure narratives more precisely.