Skip to contents

Classify patent claims as product or process

fn.patccat.Rd

Classifies U.S. patent claims as product or process and attaches claim-structure flags, using udpipe part-of-speech tags.

Usage

fn.patccat(data, varPatentClaim = "PatentClaim", varText = "text",
  varLevel = "level", varSequence = "sequence", varID = "id",
  extractPreambleTerms = TRUE, extractPreambleText = TRUE,
  timeOutput = TRUE, integrityCheck = TRUE,
  params = NULL, claim.rules = NULL, process.words = NULL,
  product.words = NULL, processwords.simple = NULL)

Arguments

data

A data frame of claim lines: one row per claim line.

varPatentClaim

Column naming the claim key (default "PatentClaim").

varText

Column with the claim text (default "text").

varLevel

Column with the line level (default "level").

varSequence

Column with the line sequence (default "sequence").

varID

Column with a unique row id (default "id").

extractPreambleTerms

Extract preamble terms (default TRUE).

extractPreambleText

Extract preamble text; required for Beauregard detection (default TRUE).

timeOutput

Print timing (default TRUE).

integrityCheck

Run the input integrity check (default TRUE).

params, claim.rules, process.words, product.words, processwords.simple

Optional overrides for the classifier's internal parameters and word lists. Each defaults to NULL, meaning the validated internal value is used, so the default call reproduces the classifier exactly. Supplying your own values changes the classification and leaves the validated AMT baseline behind; intended for research and experimentation. See patccat_defaults to inspect the internal values (if available), or the package data.

Value

A data frame with one row per independent claim, keyed by PatentClaim. The label is claimType (0 = uncategorized, 1 = process/method, 2 = product); product-by-process is folded into product and flagged in prodByProcess. Alongside it are the claim-structure flags, including independent, singleLine, Jepson, isMeans (means-plus-function), markush, beauregard, nonTransitory, computerImplemented, and stepPlusFunction, plus structural fields such as preambleType, bodyType, transitionType, and the word counts wordsPreamble and wordsBody.

Details

Requires the udpipe POS model; it is loaded automatically on first use via patccat_load_model. Overrides passed here are resolved against the validated internal defaults and used for that call only.

See also

patccat_classify, patccat_load_model, fn.benchmarking

Examples

if (FALSE) { # \dontrun{
patccat_load_model()
out <- fn.patccat(claim_lines)

## research: try a custom product-word list
out2 <- fn.patccat(claim_lines, product.words = c("apparatus", "device", "widget"))
} # }