Download a curated data frame based on episode and movie transcripts containing metadata and variables for analysis of scenes, character presence, dialog, sentiment, etc.
st_transcripts(type = c("clean", "raw"))character, "clean" for curated nested data frame or "raw" for
unprocessed text. See details.
a tibble data frame
The data frame contains metadata associated with each transcript, one row per
episode. It also contains a list column. By default (type = "clean"), this
is a nested data frame of preprocessed text split into several variables
including the speaking character, line spoken, scene descriptions, etc. For
the raw text version, the list column contains vectors of unprocessed plain
text.
Metadata includes the format (episode or movie), series, season, overall
episode number, title, production order and original airdate if available and
applicable. The two columns url and url2 show where source material can
be browsed online, though not in a useful format for data analysis. The first
set is used if possible because it contains more complete, higher quality
data. When necessary, the derived data is based on text from the alternate
source.
The dataset is nicely curated, but imperfect. There are text-parsing edge
cases that are difficult to handle generally. The quality varies
substantially across series. Datasets assembled based on original transcripts
are more informative, but not universally available. Other episodes are based
on transcripts derived from closed captioning, in which case more fields will
contain NA values.
This function downloads and returns a sizable tibble data frame. Each version is about 13-15 MB compressed. The returned tibble contains 726 rows (716 episodes and 10 movies), but each row has nested data.
if (FALSE) stTranscripts <- st_transcripts()