I'm not familiar with the output of the quanteda package but if you can conver the output to a dataframe then you can use tidyr::separate()
library(tidyr)
df <- data.frame(stringsAsFactors = FALSE,
bigram = c("good_morning", "right_now", "years_ago",
"last_night", "r_u", "ou_know"))
separate(df, bigram, c("word1", "word2"), sep = "_")
#> word1 word2
#> 1 good morning
#> 2 right now
#> 3 years ago
#> 4 last night
#> 5 r u
#> 6 ou know
Created on 2019-04-15 by the reprex package (v0.2.1.9000)