Knowledge Cartography

What you can see is a knowledge landscape calculated from 33 OER textbooks from OpenLearn (= 741 chapters) in the area of business.

Hills indicate a pile of semantically connected concepts (=keywords) and textbook chapters (=RSS items of the openlearn feed), valleys indicate their absence. There are 1193 concepts in this map.

Each corner in the map is a 'semantic corner' of closely related topics; in sum the landscape covers these 20 textbooks.

There are two types of markers put into the map:

The footprints indicate a user trail: you can imagine them to represent, what topics and textbook chapters the user has visited. If we had assessment information, we could also mark up, what ground has been covered (and what charted semantic territory still has to be covered).

The other markers (= aggregate radars and the text icons depicted to the left) indicate the 'concepts' (= keywords) and pages of the textbooks. You can click on the text icons to see whether they are a concept (e.g. 'characteristics') or a textbook chapter.

You can mouse over the user trails to see what they are: e.g. B629_1_item-5 is the 5th chapter "Introduction to Financial Stakeholders" of the textbook "Stakeholders in marketing and finance (B629_1)".

You can zoom in and out; you can click on the aggregate markers to zoom in.

This is just a demo. What is lacking for a fully scaling prototype is:

• calculate individual maps for each zoom level (right now it's only one pic cut into pieces, but we could easily calculate higher resolution ones for the lower zoom levels -- including a network path view on the lowest level)
• use different markers for textbook chapters and for concepts (= keywords)
• implement better filtering of the concepts (=keywords), so that there is only a selection of the most central concepts of a pile shown -- and not all of them.
• add interaction with openlearn, so that users can jump to the chapter
• the placement of the markers is not accurate (the mapping of the map coordinates to Google maps geocoordinates is a bit messed up); right now it is 'about right' -- but you can already see that the cluster markers are not on top of the hills (where they correctly would be).
• prevent image wrapping, so that you can't pan across the boundaries of the image.

The underlying technology is Meaningful Interaction Analysis as introduced in Wild, Haley, & Buelow (2011): Using Latent-Semantic Analysis and Network Analysis for Monitoring Conceptual Development, In: Journal for Language Technology and Computational Linguistics, 26(1):9-21.