Improving Labbé’s Intertextual Distance: Testing a Revised Version on a Large Corpus of Italian Literature

Abstract

Moving from Labbé's proposal envisaging the use of intertextual distance to measure the similarity (and dissimilarity) of texts, this paper proposes a new calculation procedure based on repeated observations of intertextual distance between pairs of equal-sized text chunks. The implementation of this procedure on a large corpus including 160 Italian novels provides information on the values produced by measuring intertextual distance in (both lemmatized and non-lemmatized) literary texts written in Italian. In order to show the improvement achieved through this iterative procedure compared to the original version, distance values are assessed in terms of their ability to recognize the author as the factor responsible for text pairing.

Towards a taxonomy of suspected forgery in authorship attribution field

“…New calibration and applications of this index are having particularly fruitful results (see e.g. [20], [54]). …”

“…A further improvement of the index offers a new calculation procedure based on repeated observations of intertextual distance between pairs of equal-sized text chunks [20]. The abovementioned study provides some tests on the thresholds proposed by Labbé aimed at considering the specificities of the Italian language.…”

“…In the previous years a poet like Giovanni Raboni, among others, had raised a doubt about the authenticity of the texts (although on the basis of an aesthetic judgment only [49]), but immediately after the publication of [46] another of the most authorative Montalian philologists, Dante Isella, launched his formidable attack against the authenticity [30]. In addition to philological arguments (based especially on the patchwork nature of many texts, which look like a mosaic of Montalian quotations: see section 4.1), Isella underlined -through an impressive expertise by Armando Petrucci, perhaps the most famous Italian paleographer -the highly implausible appearance of the supposed autographs published, in the meantime, by Annalisa Cima [30,20]). The reply by the supporters of the authenticity of Diario was immediate: those attending a Conference in Lugano, in autumn 1997 [1], where the first exhibition of the discussed autographs took place, affirmed the authenticity of the collection unanimously.…”

A chi assomiglia Elena Ferrante? Un profilo stilometrico aggior-nato

“…Sono state usate tecniche che, basandosi su un'analisi quantitativa del lessico delle opere confrontate, permettesse di leggere «da distante» (per rifarci alla nota espressione di Moretti, 2013) l'insieme delle opere e di vedere in un colpo d'occhio la collocazione reciproca di tutti gli autori e di tutte le opere oggetto del nostro studio. Le tecniche usate sono state due: un'analisi, esplorativa, condotta attraverso l'analisi delle corrispondenze (Greenacre, 1984) e un'analisi effettuata attraverso la misura della distanza intertestuale di Labbé (Labbé & Labbé, 2001 che, pur presentando alcuni limiti (Viprey & Ledoux, 2006), si è rivelata efficace negli esperimenti condotti su testi di certa attribuzione, soprattutto nella versione iterativa da noi proposta (Cortelazzo, Nadalutti & Tuzzi, 2013;Tuzzi, 2010). L'analisi delle corrispondenze applicata all'intero corpus ha fatto emergere tre autori che appaiono i più originali; tra questi Elena Ferrante (gli altri due sono Paolo Nori e Giorgio Faletti) mostra la più marcata individualità tematica e di scrittura, finendo isolata e rappresentata a notevole distanza anche dagli scrittori che evidenziano qualche affinità, come i meridionali (in gran parte napoletani) Prisco, Rea, Starnone, De Luca, Montesano, Carofiglio, il milanese, ma di famiglia meridionale, Balzano e, fuori da questo quadro, la romana Clara Sereni.…”

Measuring Global Similarity Between Texts

“…Statistical approaches for comparing texts are used for example in machine translation for assessing the quality of machine translation tools [18,19,22], or in computational linguistics in order to establish authorship [3,14,17,24,25,30] or to detect "fake", i.e., automatically generated, scientific papers [16,21].…”