smoothers: k-gram Probability Smoothers

In kgrams: Classical k-gram Language Models

k-gram Probability Smoothers

Description

Information on available k-gram continuation probability smoothers.

List of smoothers currently supported by kgrams

• "ml": Maximum Likelihood estimate \insertCitemarkov1913essaikgrams.

• "add_k": Add-k smoothing \insertCitedale1995philosophical,lidstone1920note,johnson1932probability,jeffreys1998theorykgrams.

• "abs": Absolute discounting \insertCiteney1991smoothingkgrams.

• "wb": Witten-Bell smoothing \insertCitebell1990text,witten1991zerokgrams

• "kn": Interpolated Kneser-Ney. \insertCiteKneser1995ImprovedBF,chen1999empiricalkgrams.

• "mkn": Interpolated modified Kneser-Ney. \insertCitechen1999empiricalkgrams.

• "sbo": Stupid Backoff \insertCitebrants-etal-2007-largekgrams.

Usage

smoothers()

info(smoother)

Arguments

smoother

a string. Code name of probability smoother.

Value

smoothers() returns a character vector, the list of code names of probability smoothers available in kgrams. info(smoother) returns NULL (invisibly) and prints some information on the selected smoothing technique.

Author(s)

Valerio Gherardi

References

\insertAllCited

Examples

# List available smoothers
smoothers()

# Get information on smoother "kn", i.e. Interpolated Kneser-Ney
info("kn")

kgrams documentation built on April 4, 2025, 1:10 a.m.