get_enriched_themes {stoRy}

Find over-represented themes in a collection

Description

get_enriched_themes() calculates the top m most over-represented (or enriched) themes in a sub-collection of interest from a background collection.

Usage

get_enriched_themes(
  test_collection,
  background_collection = NULL,
  top_m = 10,
  weights = list(choice = 3, major = 2, minor = 1),
  explicit = TRUE,
  min_freq = 1,
  blacklist = NULL,
  metric = c("hgt", "tfidf")
)

Arguments

test_collection

A Collection() class object of stories to assay for over-represented themes.

background_collection

A Collection() class object the stories of are a superset of the test_collection stories.

If NULL, the collection of all stories in the actively loaded LTO version is used.

top_m

Maximum number of themes to report. The default is top_m=10.

If Inf, all themes occurring at least min_occurrence times in the collection are reported.

weights

A list assigning nonnegative weights to choice, major, and minor theme levels. The default weighting list(choice = 3, major = 2, minor = 1) counts each choice usage three times, each major theme usage twice, and each minor theme usage once. Use the uniform weighting list(choice = 1, major = 1, minor = 1) weights theme usages equally regardless of level. At least one weight must be positive.

explicit

Set to FALSE to include ancestor themes of the explicit thematic annotations.

min_freq

Drop themes occurring less than this number of times from the analysis. The default min_freq=1 results in no themes are discarded.

blacklist

A Themeset() class object. A themeset containing themes to be dropped from the analysis.

If NULL, no themes are filtered.

metric

A character vector specifying the choice of scoring function. Use metric = "hgt" for the hypergeometric test, and metric = "tfidf" for term frequency-inverse document frequency. The default specification of metric = c("hgt", "tfidf") results in the hypergeometric test being used in the analysis.

Details

The test collection of n stories, S[1], \ldots, S[n], is represented as a weighted bag-of-words, where each choice theme in story S[j] (j=1, \ldots, n) is counted weights$choice times, each major theme weights$major times, and each minor theme weights$choice times.

The background collection of N stories, S[1], \ldots, S[N], is a superset of the test collection that is likewise represented as a weighted bag-of-words.

Theme enrichment scores are calculated according to the hypergeometric test by default. Set metric = "tfidf" to use TF-IDF weights for the enrichment scores.

Value

Returns a tibble with top_m rows (themes) and 10 columns:

theme_name:	m-th most over-represented theme in the test collection
k:	Number of test collection stories featuring the theme
k_bar:	Weighted counts of the theme summed over the test collection stories
n:	Number of stories in the test collection
n_bar:	Sum of all weighted counts of test collection themes
K:	Number of background collection stories featuring the theme
K_bar:	Weighted counts of the theme summed over the background collection stories
N:	Number of stories in the background collection
N_bar:	Sum of all weighted counts of background collection themes
score:	Either the negative base 10 logarithm of the Hypergeometric test (if metric = "hgt") or TF-IDF (if metric = "tfidf")

References

Mikael Onsjö, Paul Sheridan (2020). Theme Enrichment Analysis: A Statistical Test for Identifying Significantly Enriched Themes in a List of Stories with an Application to the Star Trek Television Franchise. Digital Studies/le Champ Numérique, 10(1), 1. DOI: doi:10.16995/dscn.316

Examples

## Not run: 
# Retrieve the top 10 most enriched themes in "The Twilight Zone" (1959)
# series episodes with all demo version stories as background:
set_lto("demo")
test_collection <- Collection$new(collection_id = "Collection: tvseries: The Twilight Zone (1959)")
result_tbl <- get_enriched_themes(test_collection)
result_tbl

# Run the same analysis on "The Twilight Zone" (1959) series without
# including minor level themes:
result_tbl <- get_enriched_themes(test_collection, weights = list(choice = 1, major = 1, minor = 0))
result_tbl

## End(Not run)