Računalniško podprta stilometrična analiza pripovedne literature Janeza Ciglerja in Christopha Schmida v slovenščini

Žejn, Andrejka

doi:10.31820/f.32.2.5

Cited by 1 publication

(3 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…First, we constructed five datasets consisting of comments in Slovenian, each with a different number of included authors (5,10,20,50,100). When constructing a dataset that includes the comments of N most active users by the number of comments, we consider each user as a separate class for the purpose of the AA classification task.…”

Section: Methodsmentioning

confidence: 99%

“…Panker [19] tested a variety of computational approaches based on stylometry (the most effective approach was the Naive Bayes classifier), reaching an accuracy of 75%. Žejn [20] attempted to determine whether some notable Slovenian authors truly authored all of their works, but did not attempt to generalize the method or numerically quantify the results.…”

Section: Related Workmentioning

confidence: 99%

“…Therefore, in this study, we propose an approach to tackle this task for Slovenian texts. We collect short texts from a Slovenian news website comment section and fine-tune a number of models that differentiate between a different number of authors (5,10,20,50,100). The models are based on the monolingual SloBERTa model [8] and the multilingual BERT model [9].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Authorship Attribution on Short Texts in the Slovenian Language

Gabrovšek,

Peer,

Emeršič

et al. 2023

Applied Sciences

View full text Add to dashboard Cite

The study investigates the task of authorship attribution on short texts in Slovenian using the BERT language model. Authorship attribution is the task of attributing a written text to its author, frequently using stylometry or computational techniques. We create five custom datasets for different numbers of included text authors and fine-tune two BERT models, SloBERTa and BERT Multilingual (mBERT), to evaluate their performance in closed-class and open-class problems with varying numbers of authors. Our models achieved an F1 score of approximately 0.95 when using the dataset with the comments of the top five users by the number of written comments. Training on datasets that include comments written by an increasing number of people results in models with a gradually decreasing F1 score. Including out-of-class comments in the evaluation decreases the F1 score by approximately 0.05. The study demonstrates the feasibility of using BERT models for authorship attribution in short texts in the Slovenian language.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%