To get rid of stop words in a tidy format, you can use the stop_words data frame provided in the tidytext package. You call that tibble into the environment, then do an anti-join by word:

> library(tidytext)

> data(stop_words)

> sotu_tidy <- sotu_unnest %>%
    dplyr::anti_join(stop_words, by = "word")

Notice that the length of the data went from 1.97 million observations down to 778,161. Now, you can go ahead and see the top words. I don't do it in the following, but you can put this into a data frame if you so choose: 

> sotu_tidy %>%
    dplyr::count(word, sort = TRUE)
# A tibble: 29,558 x 2
   word           n
   <chr>      <int>
 1 government  7573
 2 congress    5759
 3 united      5102
 4 people      4219
 5 country     3564
 6 public      3413
 7 time        3138
 8 war         2961
 9 american    2853
10 world       2581
# ... with 29,548 more rows

We can pass this data to ggplot2, in this case, words that occur more than 2,500 times:

> sotu_tidy %>%
    dplyr::count(word, sort = TRUE) %>%
    dplyr::filter(n > 2500) %>%
    dplyr::mutate(word = reorder(word, n)) %>%
    ggplot2::ggplot(ggplot2::aes(word, n)) +
    ggplot2::geom_col() +
    ggplot2::xlab(NULL) +
    ggplot2::coord_flip() +
    ggthemes::theme_igray()

The output of the preceding code is as follows:

We can look at the addresses that contain the most total words: 

> sotu_tidy %>%
    dplyr::group_by(year) %>%
    dplyr::summarise(totalWords = length(word)) %>%
    dplyr::arrange(desc(totalWords))
# A tibble: 225 x 2
    year totalWords
   <int>      <int>
 1 1981       18402
 2 1980       17553
 3 1946       12614
 4 1974       11813
 5 1979       11730
 6 1910       11178
 7 1907       10230
 8 1912       10215
 9 1911        9598
10 1899        9504
# ... with 215 more rows

How about that? The two longest speeches were given by Ronald Reagan, often called The Great Communicator. Moving on, we'll take a look at Lincoln's top word frequency, then create a word cloud for each of the separate addresses.