One might assume that the average adjective is more frequent than the average noun. This might be the case because there are so much nouns, consequently having a lower average frequency. If one further assumes that most NN-co-occurrences are of the type ADJ-N (or DET-N, which works the same), then on might to compare the ranks of co-occurring neighbors. This is done here and the following subsection.
First we consider the percentage of co-occurrences with more frequent left part for the right part being in the top-N-words.
The following diagram displays the table data with logarithmic x-axis.
The data might be useful for language comparison.
First line in table:
select 100, (select count(*) from co_n where w1_id>100 and w2_id>100 and 100+100>w1_id and 100+100>w2_id and w2_id>w1_id) / (select count(*) from co_n where w1_id>100 and w2_id>100 and 100+100>w1_id and 100+100>w2_id);
5.1.9.2 Skewness in NN co-occurrences II