Evaluate formex predictions against a labeled column corpus

Evaluate formex predictions against a labeled column corpus

Usage

evaluate_columns(
  columns,
  labels,
  crosswalk,
  headers = NULL,
  use_name = FALSE,
  ...
)

Arguments

columns

A list of character vectors, one per labeled column.

labels

A character vector of source-vocabulary labels, length == length(columns).

crosswalk

A data frame with columns source_label, data_type, data_subtype, data_class mapping each source label to a formex triple (data_class = NA marks an out-of-scope label).

headers

Optional character vector of real column headers (only used when use_name = TRUE).

use_name

If TRUE, predict with guess_column() (values + name); otherwise value-only via guess_data_type().

...

Passed to the underlying guesser (e.g. threshold, n_sample).

Value

A data frame with one row per column: label, gold_type, gold_class, pred_detector, pred_type, pred_class, source, in_scope, class_correct, type_correct.

See also

eval_summary()