Being able to identify the author of an unknown text is crucial. Although it is a well-studied field, it is still an open problem, since a standard approach has yet to be found. In this notebook, we propose our model for the Authorship Attribution task of PAN 2019, that focuses on cross-domain setting covering 4 different languages: French, Italian, English, and Spanish. We use n-grams of characters, words, stemmed words, and distorted text. Our model has an SVM for each feature and an ensemble architecture. Our final results outperform the baseline given by PAN in almost every problem. With this model, we reach the second place in the task with an F1-score of 68%.
Dettaglio pubblicazione
2019, CEUR Workshop Proceedings, Pages - (volume: 2380)
Cross-domain authorship attribution combining instance-based and profile-based features notebook for PAN at CLEF 2019 (04b Atto di convegno in volume)
Bacciu A., La Morgia M., Mei A., Nemmi E. N., Neri V., Stefa J.
keywords