This can be done using `dplyr` and a combination of `group_by` and aggregations functions. Something like this should work:

    library(dplyr)

    df_n <- data.frame(
      id = c(1, 2, 3),
      lemma = c("word1", "word2", "word3"),
      year = c(1970, 1971, 1972), 
      count = c(737, 767, 988)
    )

    df_n %>%
        # create column that specifies the decade
        mutate(decade = year - year %% 10) %>%
        group_by(decade, lemma) %>%
        # add up counts for duplicate words within groups specified above
        summarise(count = sum(count)) %>%
        group_by(decade) %>%
        # select top 10 records based on count within groups specified above
        top_n(10, count)