More examples of word-trees

[1: Is intl devt complex?] [2: Strong patterns]  [2.1: More examples of word trees] [Part 3: Scribes’ feedback]

Here are two more examples of the branching ideas found within all stories and just those related to GlobalGiving’s partners, at varying degrees of depth:

All 39,470 stories’ ideas branching out (click to see full version)

As this level of zoom, I’ve manually added some coloring of concept-related branches:

Important note! When I color a word, such as farming, to associate it with trees and agriculture or HIV, any adjacent words in the map also get colored by Gephi.

 The core blob of this giant set of stories is primarily people and education centered, based on colors.

Conceptual map of thousands stories related to GlobalGiving Partner Organizations

This is not an exact list of stories, since the storytellers don’t exactly know who did what, when, and with whom, in their own community. But here are the stories we THINK are related to our partners, at varying levels of branching depth. (I can generate a simplified map or an incredibly branchy map, and both are useful):

Simplest map of the ‘core’ subjects in thousands of GG partner-related stories. Each word was used at least 20 times, and there are 464 lines between nodes in this graph. A few nodes are disconnected – common words but the stories containing them did not connect to other common ideas and words.
Still shows only GlobalGiving-partner-related stories, but algorithm allows for more branching.
And the full* picture – where we allow even more branching. Words still must be used at least 10 times within these stories overall to appear somewhere in the tree. But a single set of stories (defined by containing one or more words drawn from the center of the blob) can have as many as 15 words branching off of it.

Reading the big word tree

  • At this deeper view into the story topics, you see much more interweaving of education, health, economic, and agricultural themes in stories. However, many health branches remain distinct within the tree as a whole – probably because these stories remain more single-issue focused.
  • In contrast, the education branch has disappeared into the center mass of ideas because education overlaps with everything.
  • The spread of HIV is interwoven into many stories (in the center), whereas malaria prevention remains an isolated problem.
  • Kibera (a slum in Nairobi, Kenya) has many branching problems. No other locations stand out.
  • Shallow Water International Movement (SWIM) is an organization that tried to join GlobalGiving in a past open challenge. They helped collect stories, which unfortunately, were extremely redundant and focused entirely on themselves. Hence the words Shallow, Wells, International, and Movement are all on the map as the stamp of one organization’s effort to draw attention to themselves. The sad part is that they work in a remote area with many needs, and many of the people in these places were not heard (other than the message that they need water).
  • The SWIM effect is one weakness of this approach. People may be saying the same thing but not using any of the same words. Do analyze meaning in that context would be very difficult. But there are many things there which match my expectations of what stories about organizations would be likely to cover.

Word/concept maps of the messiest type of story

Below is a map of stories where many of the elements are ‘mixed’ and not clear: Mixture of need, problem, and solution; mixture of social relations, physical well-being, and economic opportunity, and a mixture of success and failure – in the storyteller’s point of view:

MIXED need/problem/solution + MIXED social/physical/economic + MIXED outcome (failure vs success) yields a map in which nearly all the nodes are people (green) oriented.

I didn’t try to color-code nodes by whether they are related to behavior change, but I would be willing to bet many of these stories are about intractable ‘behavior change’ problems in development.  Here is a file containing all of these mixed stories so you can decide for yourself :

7300 stories with mixed outcomes and mixed elements

And if you want to download the free and open-source Gephi viz tool and play with these files, here they are. (Change the extension to .gephi if it downloads as a .doc file instead):




Part of the definition of a ‘chaotic system’ is that it is extremely sensitive to initial conditions. In these examples, the structure of the whole map is quite different depending on two parameters:

  1. Branchiness – given a set of stories that contain the same word, how many other words can the network allow to ‘grow’ off of it?
  2. Word frequency – When pulling the original set of “most common words” out at the top level, what is the minimum number of times that each word must appear across all stories to be considered for this graph? As you lower this threshold, you reveal smaller, hidden veins of ideas underneath the broader trains of thought.
  3. Pruning – the algorithm removes words shared in multiple branches using the highlander rule (there can be only one), so the one with the greater number of uses within its set of stories wins. This is why words seem to jump around within the branch under different conditions – under different conditions the word is pruned and appears in different part of the tree, yet in real life many stories contain the overlapping word, but just not as concentrated.

Eventually, with lots more time, I’d like to systematically measure how different all possible maps appear relative to each other as a function of altering branchiness and minimum word frequency. I think that map diversity provides us a proxy measure for the “messiness” or complexity of the underlying concepts described. Unfortunately, that isn’t the purpose of this project; the goal this project is to find a cheap and effective way to provide feedback to hundreds of partner organizations, so that the field of international development can operate more effectively. But it is food for thought…

Continue reading: [3: What scribes think about collecting stories]

2 thoughts on “More examples of word-trees

  1. Hi Maxson These are great works we must appreciate the Story Telling through The GlobalGiving and am sure that Western HIVAIDS Network will soon get somewhere because now our organization has been promoted to belong to the IDC – International Development Community

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s