Humanities Data Analysis “125 85018 Karsdrop Humanities ch01 3p” — 2020/8/19 — 11 05 — page 248 — #1 8CHAPTER Stylometry and the Voice of Hildegard ���������������������������������������������������[.]
“125-85018_Karsdrop_Humanities_ch01_3p” — 2020/8/19 — 11:05 — page 248 — #1 CHAPTER Stylometry and the Voice of Hildegard 8.1 Introduction Hildegard of Bingen, sometimes called the “Sybil of the Rhine,” was a famous author of Latin prose in the twelfth century Female authors were a rarity throughout the Middle Ages and her vast body of mystical writings has been the subject of numerous studies Her epistolary corpus of letters was recently investigated in a stylometric paper focusing on the authenticity of a number of dubious letters traditionally attributed to Hildegard (Kestemont, Moens, and Deploige 2015) For this purpose, Hildegard’s letters were compared with that of two well-known contemporary epistolary oeuvres: that of Bernard of Clairvaux, an influential thinker, and Guibert of Gembloux, her last secretary Figure 8.1 (reproduced from the original paper) provides a bird’s eye visualization of the differences in writing style between these three oeuvres Documents written by Hildegard, Bernard of Clairvaux, and Guibert of Gembloux are assigned the prefixes H_, B_, and G_ respectively The words printed in gray show which specific words can be thought of as characteristic for each author’s writing style As can be gleaned from the scatter plot in the left of the panel, these oeuvres fall into three remarkably clear clusters for each author, suggesting that the three authors employed markedly distinct writing styles The goal of this chapter is to reproduce (parts of) the stylometric analysis reported in Kestemont, Moens, and Deploige (2015) While chapter already briefly touched upon the topic of computational stylometry, the current chapter provides a more detailed and thorough overview of the essentials of quantitatively modeling the writing style of documents (Holmes 1994, 1998) This has been a significant topic in the computational humanities and it continues to attract much attention in this field (Siemens and Schreibman 2008) Stylometry is typically concerned with modeling the writing style of documents in relation to, or even as a function of, metadata about these documents (cf Herrmann, Dalen-Oskam, and Schöch 2015 for a discussion of the definition of style in relation to both literary and computational literary stylistics) A typical question would for instance be how the identity of a document’s author might be “125-85018_Karsdrop_Humanities_ch01_3p” — 2020/8/19 — 11:05 — page 249 — #2 Stylometry and the Voice of Hildegard • sicut 0.1 30 ut licet super H_epNG-3 per quasi cum 0.0 ne pro qui -0.1 H_epNG-6 uelut H_epNG-5 etiam H_epNG-4 propter H_epNG-8 H_epNG-2 ipse unde xque H_epNG-7 ita tunc H_epNG-1 atque ante ac quia autem nunc sic hic sed -0.2 PC2 (16.9%) post inter semper usque dum uel G_ep-7 e quoque siue scilicet ad quoniam in G_ep-3 G_ep-4 G_ep-5 G_ep-11 G_ep-6 G_ep-1 G_ep-8 G_ep-2 a G_ep-12 G_ep-9 G_ep-10 quod adhuc B_ep-7 quantum uidelicet B_ep-9 sine B_ep-13 apud B_ep-12 B_ep-14 quam de B_ep-16 aut B_ep-8 contra tam B_ep-6 nisi magis B_ep-11 nec iam tamen enim quidem B_ep-3 B_ep-15 ubi si B_ep-10 B_ep-5 ergo B_ep-2 B_ep-1 non 25 0.2 idem 20 et 15 10 -2 -4 -4 -2 Proportion of variance explained (in %) -6 35 Principal Components Analysis -6 -0.3 B_ep-4 -0.2 -0.1 0.0 0.1 PC1 (37.8%) 65 MFW Culled @ 0% Pronouns deleted Correlation matrix 0.2 -0.3 Principal components Figure 8.1 A Principal Component Analysis (PCA) plot (first dimensions) contrasting 10,000 lemmatized word samples from three oeuvres (Bernard of Clairvaux, Hildegard of Bingen, and Guibert of Gembloux) After Kestemont, Moens, and Deploige (2015) predicted from the document’s writing style—an issue which is central in the type of authorship studies of which The Federalist Papers covered in chapter is an iconic example (Mosteller and Wallace 1964) While authorship studies are undeniably the most popular application of stylometry, stylometry is rapidly expanding its scope to include other fields of stylistic inquiry as well In “Stylochronometry” (Stamou 2008), for instance, scholars attempt to (relatively or absolutely) date texts This might be a useful diachronic approach to oeuvre studies, focusing on the works of a single individual, or when studying historical texts of which the exact date of composition might be unknown or disputed Other recent applications of stylometry have targeted meta-variables related to genre (Schöch 2017), character interaction (Karsdorp, Kestemont, Schöch, and Van den Bosch 2015), literary periodization (Jannidis and Lauer 249 ... Components Analysis -6 -0.3 B_ep-4 -0.2 -0.1 0.0 0.1 PC1 (37.8%) 65 MFW Culled @ 0% Pronouns deleted Correlation matrix 0.2 -0.3 Principal components Figure 8.1 A Principal Component Analysis (PCA)...“125-85018_Karsdrop _Humanities_ ch01_3p” — 2020/8/19 — 11:05 — page 249 — #2 Stylometry and the Voice of Hildegard •