Machine Learning and Feature Selection for Authorship Attribution: The Case of Mill, Taylor Mill and Taylor, in the Nineteenth Century

Neocleous, Andreas; Loizides, Antis

doi:10.1109/access.2020.3047583

Cited by 10 publications

(12 citation statements)

References 30 publications

(25 reference statements)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…However the SVMs attribute parts of some of the chapters of "SoW" and "OL" to Harriet Taylor Mill. These results are in line with [1].…”

Section: E Test Setsupporting

confidence: 92%

“…Following the corpus in [1], we used the standard editions for John Stuart Mill's [39] and Harriet Taylor Mill's [40] authorial set. It should be noted that, unlike with John Stuart Mill, there was no much choice of texts in the case of Harriet Taylor Mill.…”

Section: ) Deviations From Previous Datasetmentioning

confidence: 99%

“…From the results in [1], we concluded that creating a four-class classification model is not the optimal way to go through. In this work we decided to simplify the models into training binary classifiers to distinguish between two authors.…”

Section: Introductionmentioning

confidence: 98%

“…In this paper we attempt to expand our previous work [1], in which we made our first attempt to use standard machine learning techniques for creating models that can distinguish the writing style of three different authors in three different classes, and at the same time, to create another class in those VOLUME 4, 2016 models which can distinguish essays of joint contributions.…”

Section: Introductionmentioning

confidence: 99%

“…Using the corpus used in [1], though without the Helen Taylor texts, we attempt to move past a methodological limitation which the sparing number of instances presents. In the process, we also examine whether the conclusion that it was John Stuart Mill who wrote the two texts in question still holds, despite the change in methodology.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Text Mining in 19th-Century Essays for Investigating a Possible Collaborative Authorship Problem: John Stuart Mill and Harriet Taylor Mill

2022

Self Cite

View full text Add to dashboard Cite

In this work, we use machine learning techniques to address a research question regarding the authorship attribution of two famous essays in the nineteenth century. On Liberty (1859) and The Subjection of Women (1869) were published under John Stuart Mill's name, a widely studied nineteenth-century British philosopher. Mill himself attributed them to collaboration with his wife and partner, Harriet Taylor Mill. More than 150 years later, the question remains whether the author of these two canonical texts in the history of political thought, was solely John Stuart Mill. Experts are divided on taking John Stuart Mill's attribution at face value, since Harriet Taylor Mill had died in 1858. Addressing this question, we use a dataset consisted in essays of both authors, to train three state-of-the-art classifiers that are able to learn and distinguish the writing style of each author. Then, we use the models built to attribute the two famous essays of disputed authorship to one of the two. From the results, we conclude that the classifiers are able to learn the two classes very well, and they return high accuracies on the validation set. Regarding the test set, most of the models attribute the two essays to John Stuart Mill, however, the contribution of Harriet Taylor Mill is shown for some chunks of text of both essays. These results, we conclude, explain why experts are divided on this particular research question.

show abstract

“…However the SVMs attribute parts of some of the chapters of "SoW" and "OL" to Harriet Taylor Mill. These results are in line with [1].…”

Section: E Test Setsupporting

confidence: 92%

Section: ) Deviations From Previous Datasetmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 98%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%