Find Nearest Tokens in Embedding Space

Usage

find_nearest(
  object,
  newdata,
  top_n = 10L,
  method = c("cosine", "euclidean", "minkowski", "dot_prod", "anchored"),
  ...,
  each = FALSE,
  include_self = TRUE,
  decreasing = NULL,
  get_sims = FALSE
)

Arguments

object: an embeddings object made by load_embeddings() or as.embeddings()
newdata: a character vector of tokens indexed by object, an embeddings object of the same dimensionality as object, or a numeric vector of the same dimensionality as object.
top_n: integer. How many nearest neighbors should be output? If each = TRUE, how many nearest neighbors should be output for each token?
method: either the name of a method to compute similarity or distance, or a function that takes two vectors and outputs a scalar, like those listed in Similarity and Distance Metrics. The value is passed to get_sims().
...: additional parameters to be passed to method function by way of get_sims()
each: logical. If FALSE (the default), the embeddings of newdata are averaged and the function will output an embeddings object with the top_n nearest tokens to the overall average embedding. If TRUE, the function will output a named list with one embeddings object for each token in newdata.
include_self: logical. Should the token(s) in newdata be included in the output?
decreasing: logical. Determines the order of sorting the similarity scores. If TRUE (the default for "cosine", "dot_prod", and "anchored" methods), the embeddings with the highest similarity scores are output. If FALSE (the default for "euclidean" and "minkowski" methods), the embeddings with the lowest distance scores are output.
get_sims: logical. If FALSE (the default), output an embeddings object or list of embeddings objects. If TRUE, output a tibble with similarity or distance scores for each document.

Value

If get_sims = FALSE (the default), an embeddings object or list of embeddings objects. If get_sims = TRUE, a tibble or list of tibbles with similarity or distance scores for each document, with columns doc_id and the name of the requested method.

Examples

words <- c("happy", "sad")
words_embeddings <- predict(glove_twitter_25d, words)

find_nearest(glove_twitter_25d, words)
#> # 25-dimensional embeddings with 10 rows
#>       dim_1 dim_2 dim_3 dim_4 dim_5 dim_6 dim_7 dim_8 dim_9 dim..      
#> miss  -0.65  0.59  0.36  0.01 -1.05 -0.02  1.34  1.46 -0.84 -0.14 ...  
#> happy -1.23  0.48  0.14 -0.03 -0.65 -0.19  2.10  1.75 -1.30 -0.32 ...  
#> love  -0.63 -0.08  0.07  0.58 -0.87 -0.15  2.23  0.99 -1.32 -0.35 ...  
#> wish  -0.52  1.12  0.49 -0.25 -1.01  0.38  2.00  0.56 -0.56  0.17 ...  
#> you   -0.42  0.33 -0.09  0.20 -0.80 -0.34  2.14  0.37 -0.94  0.24 ...  
#> thank -0.90  0.70 -0.06 -0.03 -0.77 -0.95  2.04  0.53 -1.30  0.07 ...  
#> sad    0.04 -0.19  0.44 -0.15 -0.60  0.05  1.47  0.14 -0.72  0.43 ...  
#> too   -0.47  0.40  0.12 -0.00 -0.21  0.46  1.68  0.08 -0.77  0.08 ...  
#> good  -0.54  0.60 -0.15 -0.02 -0.14  0.60  2.19  0.21 -0.52 -0.23 ...  
#> hope  -0.77  0.81  0.02 -0.23 -1.28 -0.31  1.83  0.19 -0.86 -0.23 ...  
find_nearest(glove_twitter_25d, words_embeddings) # equivalent to previous
#> # 25-dimensional embeddings with 10 rows
#>       dim_1 dim_2 dim_3 dim_4 dim_5 dim_6 dim_7 dim_8 dim_9 dim..      
#> miss  -0.65  0.59  0.36  0.01 -1.05 -0.02  1.34  1.46 -0.84 -0.14 ...  
#> happy -1.23  0.48  0.14 -0.03 -0.65 -0.19  2.10  1.75 -1.30 -0.32 ...  
#> love  -0.63 -0.08  0.07  0.58 -0.87 -0.15  2.23  0.99 -1.32 -0.35 ...  
#> wish  -0.52  1.12  0.49 -0.25 -1.01  0.38  2.00  0.56 -0.56  0.17 ...  
#> you   -0.42  0.33 -0.09  0.20 -0.80 -0.34  2.14  0.37 -0.94  0.24 ...  
#> thank -0.90  0.70 -0.06 -0.03 -0.77 -0.95  2.04  0.53 -1.30  0.07 ...  
#> sad    0.04 -0.19  0.44 -0.15 -0.60  0.05  1.47  0.14 -0.72  0.43 ...  
#> too   -0.47  0.40  0.12 -0.00 -0.21  0.46  1.68  0.08 -0.77  0.08 ...  
#> good  -0.54  0.60 -0.15 -0.02 -0.14  0.60  2.19  0.21 -0.52 -0.23 ...  
#> hope  -0.77  0.81  0.02 -0.23 -1.28 -0.31  1.83  0.19 -0.86 -0.23 ...  
find_nearest(glove_twitter_25d, words, each = TRUE)
#> $happy
#> # 25-dimensional embeddings with 10 rows
#>          dim_1 dim_2 dim_3 dim_4 dim_5 dim_6 dim_7 dim_8 dim_9      
#> happy    -1.23  0.48  0.14 -0.03 -0.65 -0.19  2.10  1.75 -1.30 ...  
#> birthday -1.46  0.50  1.00  0.17 -0.68 -0.75  1.82  1.55 -1.24 ...  
#> thank    -0.90  0.70 -0.06 -0.03 -0.77 -0.95  2.04  0.53 -1.30 ...  
#> welcome  -0.97  0.88 -0.25 -0.54 -0.55 -0.44  1.46  0.78 -0.70 ...  
#> love     -0.63 -0.08  0.07  0.58 -0.87 -0.15  2.23  0.99 -1.32 ...  
#> miss     -0.65  0.59  0.36  0.01 -1.05 -0.02  1.34  1.46 -0.84 ...  
#> hello    -0.77  0.13  0.33  0.01 -0.48 -0.50  1.86  1.06 -0.57 ...  
#> thanks   -0.80  0.82 -0.28 -0.11 -0.55 -0.72  1.62  0.98 -1.00 ...  
#> merry    -1.23  0.56  0.37  0.41 -0.56 -0.65  1.62  0.36 -1.62 ...  
#> bless    -1.15  0.81 -0.26  0.77 -0.68 -0.82  1.40  0.55 -1.23 ...  
#> 
#> $sad
#> # 25-dimensional embeddings with 10 rows
#>          dim_1 dim_2 dim_3 dim_4 dim_5 dim_6 dim_7 dim_8 dim_9      
#> sad       0.04 -0.19  0.44 -0.15 -0.60  0.05  1.47  0.14 -0.72 ...  
#> feel     -0.33  0.07  0.15 -0.17 -0.26  0.94  2.28 -0.17 -1.04 ...  
#> same      0.53  0.30  0.47 -0.02 -0.00  0.33  1.30 -0.05 -0.15 ...  
#> wrong     0.68  0.45  0.26  0.38 -0.06 -0.02  1.36 -0.64 -0.65 ...  
#> meant     0.22  0.20  0.17 -0.26 -0.68 -0.37  1.32 -0.92 -0.92 ...  
#> true      0.40  0.02 -0.63  0.04 -0.01  0.11  1.84  0.55 -0.94 ...  
#> reason    0.44  0.42 -0.00  0.33 -0.09  0.02  1.87 -0.57 -0.62 ...  
#> remember  0.12  0.77  0.56  0.20 -0.91 -0.26  1.93 -0.02 -0.87 ...  
#> i        -0.26  0.59  0.62 -0.70 -0.85 -0.23  1.05  0.07 -0.55 ...  
#> know      0.30  0.70 -0.00 -0.06 -0.69 -0.14  1.95  0.01 -0.51 ...  
#> 
find_nearest(glove_twitter_25d, words_embeddings, each = TRUE) # equivalent to previous
#> $happy
#> # 25-dimensional embeddings with 10 rows
#>          dim_1 dim_2 dim_3 dim_4 dim_5 dim_6 dim_7 dim_8 dim_9      
#> happy    -1.23  0.48  0.14 -0.03 -0.65 -0.19  2.10  1.75 -1.30 ...  
#> birthday -1.46  0.50  1.00  0.17 -0.68 -0.75  1.82  1.55 -1.24 ...  
#> thank    -0.90  0.70 -0.06 -0.03 -0.77 -0.95  2.04  0.53 -1.30 ...  
#> welcome  -0.97  0.88 -0.25 -0.54 -0.55 -0.44  1.46  0.78 -0.70 ...  
#> love     -0.63 -0.08  0.07  0.58 -0.87 -0.15  2.23  0.99 -1.32 ...  
#> miss     -0.65  0.59  0.36  0.01 -1.05 -0.02  1.34  1.46 -0.84 ...  
#> hello    -0.77  0.13  0.33  0.01 -0.48 -0.50  1.86  1.06 -0.57 ...  
#> thanks   -0.80  0.82 -0.28 -0.11 -0.55 -0.72  1.62  0.98 -1.00 ...  
#> merry    -1.23  0.56  0.37  0.41 -0.56 -0.65  1.62  0.36 -1.62 ...  
#> bless    -1.15  0.81 -0.26  0.77 -0.68 -0.82  1.40  0.55 -1.23 ...  
#> 
#> $sad
#> # 25-dimensional embeddings with 10 rows
#>          dim_1 dim_2 dim_3 dim_4 dim_5 dim_6 dim_7 dim_8 dim_9      
#> sad       0.04 -0.19  0.44 -0.15 -0.60  0.05  1.47  0.14 -0.72 ...  
#> feel     -0.33  0.07  0.15 -0.17 -0.26  0.94  2.28 -0.17 -1.04 ...  
#> same      0.53  0.30  0.47 -0.02 -0.00  0.33  1.30 -0.05 -0.15 ...  
#> wrong     0.68  0.45  0.26  0.38 -0.06 -0.02  1.36 -0.64 -0.65 ...  
#> meant     0.22  0.20  0.17 -0.26 -0.68 -0.37  1.32 -0.92 -0.92 ...  
#> true      0.40  0.02 -0.63  0.04 -0.01  0.11  1.84  0.55 -0.94 ...  
#> reason    0.44  0.42 -0.00  0.33 -0.09  0.02  1.87 -0.57 -0.62 ...  
#> remember  0.12  0.77  0.56  0.20 -0.91 -0.26  1.93 -0.02 -0.87 ...  
#> i        -0.26  0.59  0.62 -0.70 -0.85 -0.23  1.05  0.07 -0.55 ...  
#> know      0.30  0.70 -0.00 -0.06 -0.69 -0.14  1.95  0.01 -0.51 ...  
#> 

rand_vec <- rnorm(25)
find_nearest(glove_twitter_25d, rand_vec)
#> # 25-dimensional embeddings with 10 rows
#>               dim_1 dim_2 dim_3 dim_4 dim_5 dim_6 dim_7 dim_8      
#> entertaining   0.23  0.27  0.07 -0.65  0.83  1.09  1.24 -1.42 ...  
#> nervous       -0.45  0.17  0.21 -0.80  0.24  0.89  1.91 -0.57 ...  
#> relaxed       -1.93 -0.55 -0.63 -0.43  0.51  2.03  0.29 -1.48 ...  
#> enjoyable     -0.66  0.07 -0.23 -0.63  0.48  1.29  1.63 -1.73 ...  
#> surprisingly  -0.62 -0.56  0.19 -1.03  0.09  1.68  0.99 -1.63 ...  
#> instructors   -0.74  0.91 -0.14 -1.04  0.85 -0.20 -0.08 -1.92 ...  
#> routines      -0.21  0.42 -0.04 -0.30  1.11  0.45  0.69 -2.46 ...  
#> productive    -0.85  0.68 -0.40 -0.81  0.97  1.20  1.38 -1.74 ...  
#> philosophical -0.43  0.11 -0.77  0.93  1.00  1.50  1.02 -1.77 ...  
#> sgt            0.47 -0.89 -0.43 -1.23 -0.24  2.14 -0.43  1.63 ...