Skip to contents

Cosine similarity of pairs of items

Source: R/pairwise_similarity.R
pairwise_similarity.Rd

Compute cosine similarity of all pairs of items in a tidy table.

Usage

pairwise_similarity(tbl, item, feature, value, ...)

pairwise_similarity_(tbl, item, feature, value, ...)

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

...

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

Details

This function uses sparse matrices internally for efficiency, which means pairs with zero similarity are excluded from the output. To include zero-similarity pairs, use tidyr::complete() on your result: complete(item1, item2, fill = list(similarity = 0)).

See also

squarely()

Examples


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

# Comparing Jane Austen novels
austen_words <- austen_books() |>
  unnest_tokens(word, text) |>
  anti_join(stop_words, by = "word") |>
  count(book, word) |>
  ungroup()

# closest books to each other
closest <- austen_words |>
  pairwise_similarity(book, word, n) |>
  arrange(desc(similarity))

closest
#> # A tibble: 30 × 3
#>    item1               item2               similarity
#>    <fct>               <fct>                    <dbl>
#>  1 Northanger Abbey    Pride & Prejudice        0.509
#>  2 Pride & Prejudice   Northanger Abbey         0.509
#>  3 Emma                Pride & Prejudice        0.493
#>  4 Pride & Prejudice   Emma                     0.493
#>  5 Mansfield Park      Pride & Prejudice        0.483
#>  6 Pride & Prejudice   Mansfield Park           0.483
#>  7 Northanger Abbey    Emma                     0.480
#>  8 Emma                Northanger Abbey         0.480
#>  9 Pride & Prejudice   Sense & Sensibility      0.479
#> 10 Sense & Sensibility Pride & Prejudice        0.479
#> # ℹ 20 more rows

closest |>
  filter(item1 == "Emma")
#> # A tibble: 5 × 3
#>   item1 item2               similarity
#>   <fct> <fct>                    <dbl>
#> 1 Emma  Pride & Prejudice        0.493
#> 2 Emma  Northanger Abbey         0.480
#> 3 Emma  Mansfield Park           0.473
#> 4 Emma  Sense & Sensibility      0.434
#> 5 Emma  Persuasion               0.410