I have list of genes with GO labels. Some of the GOs are quite general, others very specific. I believe it could influence on GO enrichment analysis that i'm going to conduct on those gene list. Now i'm looking for a way to normalize my GOs so that after normalization i could get the same gene list, but  all GOs will have approximately similar level of depth on GO tree (the same level of specificity). I think that with GOSim or GOSemSIm bioconductor packages possible somehow to manage it, but i still can not find a direct way how to figure out it. Do you have any suggestions ?

If you want to take into account the depth of a GO term you can use topGO algorithms (specially weight01) which is a package for performing GO enrichment analysis..

With GOSemSim and GOSim you could calculate the similarity between those GO and somehow take that into account when analyzing the enrichment.