Skip to contents

Delta measure of pairs of documents

Source: R/pairwise_delta.R
pairwise_delta.Rd

Compute the delta distances (from its two variants) of all pairs of documents in a tidy table.

Usage

pairwise_delta(tbl, item, feature, value, method = "burrows", ...)

pairwise_delta_(tbl, item, feature, value, method = "burrows", ...)

Arguments

tbl

Table

item

Item to compare; will end up in item1 and item2 columns

feature

Column describing the feature that links one item to others

value

Value

method

Distance measure to be used; see dist()

...

Extra arguments passed on to squarely(), such as diag and upper

See also

squarely()

Examples


library(janeaustenr)
library(dplyr)
library(tidytext)

# closest documents in terms of 1000 most frequent words
closest <- austen_books() %>%
  unnest_tokens(word, text) %>%
  count(book, word) %>%
  top_n(1000, n) %>%
  pairwise_delta(book, word, n, method = "burrows") %>%
  arrange(delta)

closest
#> # A tibble: 30 × 3
#>    item1               item2               delta
#>    <fct>               <fct>               <dbl>
#>  1 Persuasion          Northanger Abbey    0.408
#>  2 Northanger Abbey    Persuasion          0.408
#>  3 Pride & Prejudice   Sense & Sensibility 0.491
#>  4 Sense & Sensibility Pride & Prejudice   0.491
#>  5 Persuasion          Pride & Prejudice   0.769
#>  6 Pride & Prejudice   Persuasion          0.769
#>  7 Northanger Abbey    Pride & Prejudice   0.779
#>  8 Pride & Prejudice   Northanger Abbey    0.779
#>  9 Persuasion          Sense & Sensibility 0.814
#> 10 Sense & Sensibility Persuasion          0.814
#> # … with 20 more rows

closest %>%
  filter(item1 == "Pride & Prejudice")
#> # A tibble: 5 × 3
#>   item1             item2               delta
#>   <fct>             <fct>               <dbl>
#> 1 Pride & Prejudice Sense & Sensibility 0.491
#> 2 Pride & Prejudice Persuasion          0.769
#> 3 Pride & Prejudice Northanger Abbey    0.779
#> 4 Pride & Prejudice Mansfield Park      1.04 
#> 5 Pride & Prejudice Emma                1.09 

# to remove duplicates, use upper = FALSE
closest <- austen_books() %>%
  unnest_tokens(word, text) %>%
  count(book, word) %>%
  top_n(1000, n) %>%
  pairwise_delta(book, word, n, method = "burrows", upper = FALSE) %>%
  arrange(delta)

# Can also use Argamon's Linear Delta
closest <- austen_books() %>%
  unnest_tokens(word, text) %>%
  count(book, word) %>%
  top_n(1000, n) %>%
  pairwise_delta(book, word, n, method = "argamon", upper = FALSE) %>%
  arrange(delta)