Glossary » similarity matching
a technique for identify conceptual categories. Users are given a set of topics, and they are asked to rate the similarity of pairs of topics (on some scale, e.g. 1 to 10). These ratings can them be combined with a statistical technique called cluster analysis to identify groupings of topics that most closely match user ratings. However, the technique doesn’t identify the label or “meaning” of those groups, so the person doing the analysis must interpret the groups and suggest labels if they are needed (or go back to users and ask for candidate labels).
This technique can be used, for instance, to discover possible organizations of content for a website. It has the advantage over card sort techniques that the data from each user is somewhat richer (and therefore more informative), but it is generally more time consuming than card sorts.