Is there a way in R to find a combination of words (or sentences) within a certain range in a string

Question

I'm trying to find all strings with a combination of words/sentences with other words separating them but with a fixed limit.

Example : I want the combination of "bought" and "watch" but with, at maximum, 2 words separating them.

• I bought a beautiful and shiny watch -> not ok because there is 4 words between "bought" and "watch" ("a beautiful and shiny")
• I bought a shiny watch -> ok because there is 2 words between "bought" and "watch" ("a shiny")

I haven't found anything close to what I wanted on R.

To find simple words/sentences in strings I'm using str_extract_all from stringr as here :

my_analysis <- str_c("\\b(", str_c(my_list_of_words_and_sentences, collapse="|"), ")\\b")
df$words_and_sentences_found <- str_extract_all(df$my_strings, my_analysis)

Answer 1

You can use skip-grams for this:

library(tidyverse)
library(tidytext)

df <- tibble(id = 1:3,
             txt = c("I bought a beautiful and shiny watch", 
                     "I bought a shiny watch", 
                     "The watch is very shiny"))

tidy_ngrams <- df %>%
  ## use k for the skip, and n for what degree of n-gram:
  unnest_tokens(ngram, txt, token = "skip_ngrams", n_min = 2, n = 2, k = 2) 

tidy_ngrams
#> # A tibble: 33 × 2
#>       id ngram           
#>    <int> <chr>           
#>  1     1 i bought        
#>  2     1 i a             
#>  3     1 i beautiful     
#>  4     1 bought a        
#>  5     1 bought beautiful
#>  6     1 bought and      
#>  7     1 a beautiful     
#>  8     1 a and           
#>  9     1 a shiny         
#> 10     1 beautiful and   
#> # … with 23 more rows

tidy_ngrams %>%
  filter(ngram == "bought watch")
#> # A tibble: 1 × 2
#>      id ngram       
#>   <int> <chr>       
#> 1     2 bought watch

^{Created on 2022-06-03 by the reprex package (v2.0.1)}