I agree with Carl Witthoft's advice and and would recommend you to use the combn
function.
Here is my try:
# Setup articles
articles <- paste("article_", seq(1:200))
# Setup unique coder probabilities
coders <- paste0("coder_", seq(1:7))
unique.coder.combinations <- do.call(paste, c(data.frame(t(combn(coders, 2)))))
# Assignment
coder.selection <- cbind(articles, coders=sample(unique.coder.combinations, 200, replace=TRUE))
# head(coder.selection)
# > head(coder.selection)
# articles coders
# [1,] "article_ 1" "coder_1 coder_5"
# [2,] "article_ 2" "coder_2 coder_4"
# [3,] "article_ 3" "coder_4 coder_6"
# [4,] "article_ 4" "coder_3 coder_7"
# [5,] "article_ 5" "coder_5 coder_7"
# [6,] "article_ 6" "coder_3 coder_4"