Establishing Relevance of Characteristic Features for Authorship Attribution with ANN
Authorship attribution is perceived as a task of the paramount importance within stylometric analysis of texts. It encompasses author characterisation and comparison, and by observation and recognition of patterns in individual stylistic traits enables confirmation or rejection of authorship claims. Stylometry requires reliable textual descriptors and knowledge about their relevance for the case under study. One of the possible ways to evaluate this relevance is to employ a feature selection and reduction algorithm in the wrapper model. The paper presents research on such procedure applied to artificial neural networks used to categorise literary texts with respect to their authors, with importance of attributes discovered through sequential backward search.
KeywordsStylometry Authorship Attribution Characteristic Feature Feature Relevance Feature Selection Sequential Backward Search
Unable to display preview. Download preview PDF.
- 1.Argamon, S., Burns, K., Dubnov, S. (eds.): The structure of style: Algorithmic approaches to understanding manner and meaning. Springer, Berlin (2010)Google Scholar
- 2.Burrows, J.: Textual analysis. In: Schreibman, S., Siemens, R., Unsworth, J. (eds.) A Companion to Digital Humanities. Blackwell, Oxford (2004)Google Scholar
- 3.Craig, H.: Stylistic analysis and authorship studies. In: Schreibman, S., Siemens, R., Unsworth, J. (eds.) A Companion to Digital Humanities. Blackwell, Oxford (2004)Google Scholar
- 4.Fiesler, E., Beale, R.: Handbook of neural computation. Oxford University Press (1997)Google Scholar
- 6.Kavzoglu, T., Mather, P.: Assessing artificial neural network pruning algorithms. In: Proceedings of the 24th Annual Conference and Exhibition of the Remote Sensing Society, Greenwich, UK, pp. 603–609 (2011)Google Scholar