USE OF A GENERAL GRAPH-DRAWING ALGORITHM IN THE CONSTRUCTION OF ASSOCIATION MAPS

The aim of the research reported here is to investigate one possible approach to automatically generating concept association maps that might be useful to abstractors. The approach involves using stems as concept surrogates, cooccurrence to define concept links, and a general graph-drawing algorithm to position the stems in two dimensions. Some selection of links is necessary to avoid algorithm failure. Two methods of link selection are evaluated, one (Overall) giving the overall strongest links and the other (Per-stem) the strongest links for each stem. On bringing more-associated stems together and separating less-associated stems, the graph-drawing algorithm's performance is similar for both link selection methods. It is also respectable in comparison with that of a two-dimensional scaling algorithm. The overall method is better than the Per-stem at matching associations that would actually be used in abstracting, but, on the other hand, tends to produce more unconnected stems and more link crossings.