imposters.optimize: Tuning Parameters for the Imposters Method

In computationalstylistics/stylo: Stylometric Multivariate Analyses

Tuning Parameters for the Imposters Method

Description

A function to optimize hyperparameters used in the General Imposters method (see link{imposters} for further details). Using a grid search approach, it tries to define a grey area where the attribution scores are not reliable.

Usage

imposters.optimize(reference.set,
                   classes.reference.set = NULL,
                   parameter.incr = 0.01,
                   ...)

Arguments

reference.set	a table containing frequencies/counts for several variables – e.g. most frequent words – across a number of texts written by different authors. Usually, it is a corpus of known authors (at least two tests per author) that is used to tune the optimal hyperparameters for the imposters method. Such a tuning involves a leave-one-out procedure of identifying a gray area when the results returned by the classifier are not particularly reliable. E.g., if one gets 0.39 and 0.55 as the parameters, one would assume that any results of the imposters() function that lay within this range should be claimed unreliable. Make sure that the rows contain samples, and the columns – variables (words, n-grams, or whatever needs to be analyzed).
classes.reference.set	a vector containing class identifiers for the reference set. When missing, the row names of the set table will be used; the assumed classes are the strings of characters followed by the first underscore. Consider the following example: c("Sterne_Tristram", "Sterne_Sentimental", "Fielding_Tom", ...), where the classes are the authors' names. Note that only the part up to the first underscore in the sample's name will be included in the class label.
parameter.incr	the procedure tries to optimize the hyperparameters via a grid search – this means that it tests the range of values between 0 and 1 incremented by a certain fraction. If this is set to 0.01 (default), it test 0, 0.01, 0.02, 0.03, ...
...	any other argument that can be passed to the classifier; see perform.delta for the parameters to be tweaked. In the current version of the function, only the distance measure used for computing similarities between texts can be set. Available options so far: "delta" (Burrows's Delta, default), "argamon" (Argamon's Linear Delta), "eder" (Eder's Delta), "simple" (Eder's Simple Distance), "canberra" (Canberra Distance), "manhattan" (Manhattan Distance), "euclidean" (Euclidean Distance), "cosine" (Cosine Distance), "wurzburg" (Cosine Delta), "minmax" (Minmax Distance, also known as the Ruzicka measure).

Value

The function returns two scores: the P1 and P2 values.

Author(s)

Maciej Eder

References

Koppel, M. , and Winter, Y. (2014). Determining if two documents are written by the same author. "Journal of the Association for Information Science and Technology", 65(1): 178-187.

Kestemont, M., Stover, J., Koppel, M., Karsdorp, F. and Daelemans, W. (2016). Authenticating the writings of Julius Caesar. "Expert Systems With Applications", 63: 86-96.

See Also

imposters

Examples

## Not run: 
# activating a dummy dataset, in our case: Harper Lee and her Southern colleagues
data(lee)

# running the imposters method against all the remaining authorial classes
imposters.optimize(lee)

## End(Not run)

computationalstylistics/stylo documentation built on April 7, 2024, 4:12 p.m.