Skip to contents

Read and write entity patterns

Source: R/entity-patterns.R
entity_patterns_read.Rd

Read entity patterns from a file, or write them to one, for instance to keep them with a project, edit them by hand, or share them with users of spaCy in Python.

Usage

entity_patterns_read(file, label = NULL, case_insensitive = FALSE)

entity_patterns_write(x, file)

Arguments

file

the path of the file to read or write. Its extension sets the format, and must be .jsonl for writing:

  • .jsonl: spaCy's own format, one pattern per line, each a JSON object with the keys label, pattern and optionally id, as written by spaCy's EntityRuler.to_disk();

  • .json: a JSON array of the same patterns;

  • .txt (reading only): one phrase per line, all with the label given by label. Empty lines are skipped.

label

character; for a .txt file, the label of its phrases.

case_insensitive

logical; for phrases, whether they match regardless of case. See as_entity_patterns().

x

patterns to write: a spacyr_entity_patterns object, or anything that as_entity_patterns() accepts.

Value

entity_patterns_read() returns a spacyr_entity_patterns object. entity_patterns_write() returns file, invisibly.

Details

Reading and writing JSON needs the jsonlite package.

entity_patterns_write() writes JSONL only, so file must end in .jsonl. It checks the patterns before opening file, so an existing file is left as it was if they are not valid. A phrase that matches regardless of case is written with the extra key "case_insensitive": true, which entity_patterns_read() reads back. spaCy itself ignores that key, and matches the phrase exactly.

Patterns written and read back are identical to the originals.

See also

as_entity_patterns()

Examples

# \donttest{
if (requireNamespace("jsonlite", quietly = TRUE)) {
  pats <- as_entity_patterns(list(
    list(label = "PERSON",
         pattern = list(list(LOWER = list(IN = c("dr", "dr."))),
                        list(ENT_TYPE = "PERSON", ENT_IOB = "B"),
                        list(ENT_TYPE = "PERSON", ENT_IOB = "I", OP = "*"))),
    list(label = "ORG", pattern = "Acme Widgets", id = "acme")
  ))
  file <- tempfile(fileext = ".jsonl")
  entity_patterns_write(pats, file)
  readLines(file)
  identical(entity_patterns_read(file), pats)
}
#> [1] TRUE
# }