Por la indignidad del texto: medición de la robustez estilométrica en la atribución de autoría bajo corrupción textual
Álvaro Cuéllar · Arte Nuevo · 2026
The article addresses the dilemma between the philological demand for reliable texts in the digital environment and the practical need to work at scale with imperfect materials in Spanish Golden Age theatre. As a case study, it focuses on authorship attribution via stylometric methods and proposes a controlled experiment of progressive synthetic corruption on three corpora of secure authorship, in order to measure the degree of textual indignity a corpus can withstand before attribution performance deteriorates significantly. The methodology applies a standard classifier trained on relative frequencies of the most frequent words and evaluates performance as an increasing proportion of the text is replaced with external material. The results show substantial robustness across a wide degradation range and a gradual decline as corruption becomes dominant. The conclusion argues for methodological coexistence: preserving textual dignity as an editorial horizon while acknowledging that, under controlled conditions and with appropriate cautions, provisional texts can still support empirically grounded inferences of value to philology.