To run the State of the Union demo app, run the following in the R console:
To run an alternative State of the Union app, organised by decade rather than by president, run the following in the R console:
These two apps are created with the code below. The State of the Union texts and metadata are accessed through the sotu package.
Loading packages
Creating data frame
## Merge data from 'sotu' package into one df
df <- sotu::sotu_meta
df$Text <- sotu::sotu_text %>%
stringr::str_trim()
## Avoid clutter in corpus plot
# A. Distinguish between non-consecutive terms
df$president[97:100] <- "Grover Cleveland 1"
df$president[105:108] <- "Grover Cleveland 2"
# B. Get correct order of rows in data frame
# in the cases where incumbent holds a final sotu before leaving office,
# resulting in two sotus in one year
presidents <- unique(df$president)
df$president <- factor(df$president, levels = presidents)
df <- df[order(df$president),]
df$president <- as.character(df$president)
## Add decade variable for variation of app
df$decade <- stringr::str_sub(df$year, 1, 3) %>%
paste0("0s")
# And add variable for informative document tab title in that app variation
df$for_tab_title <- paste(df$president, df$year)Create ‘corporaexplorerobject’
corpus <- prepare_data(
df, # the data frame created above
date_based_corpus = FALSE, # dates are not the organising principle in the corpus
grouping_variable = "president", # group the sotu addresses by president
# The remaining arguments are not strictly necessary, but we use them to fine-tune
# how the corpus will be presented in the app
within_group_identifier = "year", # The tab header in document view will then be e.g.
# "Theodore Roosevelt – 1901"
columns_doc_info = # metadata to be included in a "Document info" tab,
colnames(df)[1:5], # in this case the first five columns in the data frame
tile_length_range = c(2, 10), # fine-tuning the length of the tiles representing
# the length of the addrsses
use_matrix = FALSE # we don't create a document term matrix, as the corpus
# is very small and searches will be fast anyway
)Run app
explore(corpus)By just changing two arguments (or even one), we create an app with a quite different organisation of the texts.
corpus <- prepare_data(
df,
date_based_corpus = FALSE,
grouping_variable = "decade", # change grouping variable
within_group_identifier = "for_tab_title", # adjust tab header in document view
columns_doc_info =
colnames(df)[1:5],
tile_length_range = c(2, 10),
use_matrix = FALSE
)
explore(corpus)See the documentation for explore() for all runtime
options.
Example 1: Tile length By default, the length of the tiles representing documents in the app have varying lengths, depending on document length. For all tiles to of the same length:
Example 2: Plot colours To change the use of colours in the corpus map, use e.g.:
Some explore() calls with pre-filled sidebar input: