label: Manually label data

In rijpma/capelinker: Machine Learning-based Record Linkage for Historical South Africa

Manually label data

Description

label wraps a sampling and the candidates function to make manual labelling of training data easier

Usage

label(
  dat_from,
  dat_to,
  persid_from,
  persid_to,
  blockvariable,
  blocktype,
  N,
  path,
  ...
)

Arguments

dat_from	a data.table
dat_to	a data.table
persid_from	string identifying the person id variable
persid_to	string identifying the person id variable
blockvariable	string identifying the blocking variable
N	the number of unique observations of the blocking varaible to be labelled, defaults to 500
...	passed to candidates for customised blocking

Details

label takes a random sample from dat_from, gathers candidates from dat_to and presents them to the user to select the match or tell that there is no match

The labelling session is interactive, and the user is presented with a choice between

• PersidOne of the numbers of persid_to

• None

At some point a "Back" option might be added

After selecting there is an annotation step, that can be done

• Cancel

• Sure

• Maybe

• Doubtful

• Ambiguous

Value

A list containing candidate pairs to be labelled

Examples

d1 = data.table::data.table(mlast = c("jong", "smid"), mfirst = c("Jan", "Jan"), wfirst = NA, wlast = NA, settlerchildren = NA, persid = c(1:2))
d2 = data.table::data.table(mlast = c("jongh", "jong", "smit"), mfirst = c("Jan", "Dirk", "Johan"), wlast = NA, wfirst = NA, settlerchildren = NA, persid = c(1:3))
label(d1, d2, "persid", "persid", "mlast", "bigram distance", 2)

rijpma/capelinker documentation built on Nov. 7, 2024, 3:06 a.m.