Calculate semantic similarity for multiple dyads — semantic_sim

This function calculates semantic similarity over a sequence of conversation exchanges for multiple dyads.

Usage

semantic_sim_dyads(conversations, method = "tfidf", window_size = 3, ...)

Arguments

conversations: A data frame with columns 'dyad_id', 'speaker', and 'processed_text'
method: A character string specifying the method to use: "tfidf", "word2vec", or "glove"
window_size: An integer specifying the size of the sliding window
...: Additional arguments passed to semantic_similarity

Value

A list containing the sequence of similarities for each dyad and the overall average similarity

Examples

library(lme4)
convs <- data.frame(
  dyad_id = c(1, 1, 1, 1, 2, 2, 2, 2),
  speaker = c("A", "B", "A", "B", "C", "D", "C", "D"),
  processed_text = c("i love pizza", "me too favorite food",
                     "whats your favorite topping", "enjoy pepperoni mushrooms",
                     "i prefer pasta", "pasta delicious like spaghetti carbonara",
                     "ever tried making home", "yes quite easy make")
)
semantic_sim_dyads(convs, method = "tfidf", window_size = 2)
#> Warning: The 'tfidf' method may not provide highly meaningful results for short conversations or those with little vocabulary overlap. Consider using 'word2vec' or 'glove' methods for more robust results.
#> Warning: The 'tfidf' method may not provide highly meaningful results for short conversations or those with little vocabulary overlap. Consider using 'word2vec' or 'glove' methods for more robust results.
#> Warning: The 'tfidf' method may not provide highly meaningful results for short conversations or those with little vocabulary overlap. Consider using 'word2vec' or 'glove' methods for more robust results.
#> Warning: The 'tfidf' method may not provide highly meaningful results for short conversations or those with little vocabulary overlap. Consider using 'word2vec' or 'glove' methods for more robust results.
#> Warning: The 'tfidf' method may not provide highly meaningful results for short conversations or those with little vocabulary overlap. Consider using 'word2vec' or 'glove' methods for more robust results.
#> Warning: The 'tfidf' method may not provide highly meaningful results for short conversations or those with little vocabulary overlap. Consider using 'word2vec' or 'glove' methods for more robust results.
#> Warning: Problem with Hessian check (infinite or missing values?)
#> $similarities_by_dyad
#> $similarities_by_dyad$`1`
#> [1] 0.5 0.5 0.5
#> 
#> $similarities_by_dyad$`2`
#> [1] 0.5 0.5 0.5
#> 
#> 
#> $overall_average
#> (Intercept) 
#>         0.5 
#>