Is writing style predictive of scientific fraud?
Publikation: Bidrag til bog/antologi/rapport › Konferencebidrag i proceedings › Forskning › fagfællebedømt
Standard
Is writing style predictive of scientific fraud? / Braud, Chloé Elodie; Søgaard, Anders.
Proceedings of the Workshop on Stylistic VariationAssociation for Computational Linguistics. Association for Computational Linguistics, 2017. s. 37-42.Publikation: Bidrag til bog/antologi/rapport › Konferencebidrag i proceedings › Forskning › fagfællebedømt
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - GEN
T1 - Is writing style predictive of scientific fraud?
AU - Braud, Chloé Elodie
AU - Søgaard, Anders
PY - 2017
Y1 - 2017
N2 - The problem of detecting scientific fraudusing machine learning was recently introduced,with initial, positive results froma model taking into account various generalindicators. The results seem to suggestthat writing style is predictive of scientificfraud. We revisit these initial experiments,and show that the leave-one-outtesting procedure they used likely leads toa slight over-estimate of the predictability,but also that simple models can outperformtheir proposed model by some margin.We go on to explore more abstractlinguistic features, such as linguistic complexityand discourse structure, only to obtainnegative results. Upon analyzing ourmodels, we do see some interesting patterns,though: Scientific fraud, for examples,contains less comparison, as well asdifferent types of hedging and ways of presentinglogical reasoning.
AB - The problem of detecting scientific fraudusing machine learning was recently introduced,with initial, positive results froma model taking into account various generalindicators. The results seem to suggestthat writing style is predictive of scientificfraud. We revisit these initial experiments,and show that the leave-one-outtesting procedure they used likely leads toa slight over-estimate of the predictability,but also that simple models can outperformtheir proposed model by some margin.We go on to explore more abstractlinguistic features, such as linguistic complexityand discourse structure, only to obtainnegative results. Upon analyzing ourmodels, we do see some interesting patterns,though: Scientific fraud, for examples,contains less comparison, as well asdifferent types of hedging and ways of presentinglogical reasoning.
M3 - Article in proceedings
SN - 978-1-945626-99-9
SP - 37
EP - 42
BT - Proceedings of the Workshop on Stylistic VariationAssociation for Computational Linguistics
PB - Association for Computational Linguistics
T2 - Workshop on Stylistic Variation
Y2 - 8 September 2017 through 8 September 2017
ER -
ID: 195014790