Some Experimental Results on Randomly Generating Formal Contexts

We investigate different simple approaches to generate random formal contexts. To this end, we consider for each approach the empirical correlation between the number of intents and pseudo-intents. We compare the results of these experiments with corresponding observations on real-world use-cases. This comparison yields huge differences between artificially generated and real-world data sets, indicating that using randomly generated formal contexts for applications such as benchmarking may not necessarily be meaningful. In doing so, we additionally show that the previously observed phenomenon of the “Stegosaurus” does not express a real correlation between intents and pseudo-intents, but is an artifact of the way random contexts are generated.