In ngmarchant/comparator: Comparison Functions for Clustering and Record Linkage

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)
library(comparator)

comparator: Comparison Functions for Clustering and Record Linkage

comparator implements comparison functions for clustering and record linkage applications. It includes functions for comparing strings, sequences and numeric vectors. Where possible, comparators are implemented in C/C++ to ensure fast performance.

Supported comparators

String comparators:

Edit-based:

• Levenshtein(): Levenshtein distance/similarity
• DamerauLevenshtein() Damerau-Levenshtein distance/similarity
• Hamming(): Hamming distance/similarity
• OSA(): Optimal String Alignment distance/similarity
• LCS(): Longest Common Subsequence distance/similarity
• Jaro(): Jaro distance/similarity
• JaroWinkler(): Jaro-Winkler distance/similarity

Token-based:

Not yet implemented.

Hybrid token-character:

• MongeElkan(): Monge-Elkan similarity
• FuzzyTokenSet(): Fuzzy Token Set distance

Other:

• InVocabulary(): Compares strings using a reference vocabulary. Useful for comparing names.
• Lookup(): Retrieves distances/similarities from a lookup table
• BinaryComp(): Compares strings based on whether they agree/disagree exactly.

Numeric comparators:

• Euclidean(): Euclidean (L-2) distance
• Manhattan(): Manhattan (L-1) distance
• Chebyshev(): Chebyshev (L-∞) distance
• Minkowski(): Minkowski (L-p) distance

Installation

You can install the latest release from CRAN by entering:

install.packages("comparator")

The development version can be installed from GitHub using devtools:

# install.packages("devtools")
devtools::install_github("ngmarchant/comparator")

Example

A comparator is instantiated by calling its constructor function. For example, we can instantiate a Levenshtein similarity comparator that ignores differences in upper/lowercase characters as follows:

comparator <- Levenshtein(similarity = TRUE, normalize = TRUE, ignore_case = TRUE)

We can apply the comparator to character vectors element-wise as follows:

x <- c("John Doe", "Jane Doe")
y <- c("jonathon doe", "jane doe")
elementwise(comparator, x, y)

# shorthand for above
comparator(x, y)

This comparator is also defined on sequences:

x_seq <- list(c(1, 2, 1, 1), c(1, 2, 3, 4))
y_seq <- list(c(4, 3, 2, 1), c(1, 2, 3, 1))
elementwise(comparator, x_seq, y_seq)

# shorthand for above
comparator(x_seq, y_seq)

Pairwise comparisons are also supported using the following syntax:

# compare each string in x with each string in y and return a similarity matrix
pairwise(comparator, x, y, return_matrix = TRUE)

# compare the strings in x pairwise and return a similarity matrix
pairwise(comparator, x, return_matrix = TRUE)

ngmarchant/comparator documentation built on March 30, 2022, 2:51 p.m.