Computational Stylistics Research Articles

Cross-linguistic authorship attribution and gender profiling. Machine translation as a method for bridging the language gap

Abstract This study explores the feasibility of cross-linguistic authorship attribution and the author’s gender identification using Machine Translation (MT). Computational stylistics experiments were conducted on a Greek blog corpus translated into English using Google’s Neural MT. A Random Forest algorithm was employed for authorship and gender profiling, using different feature groups [Author’s Multilevel N-gram Profiles, quantitative linguistics (QL), and cross-lingual word embeddings (CLWE)] in both original and translated texts. Results indicate that MT is a viable method for converting a multilingual corpus into one language for authorship attribution and gender profiling research, with considerable accuracy when training and testing datasets use identical language. In the pure cross-linguistic scenario, higher accuracies than the baselines were obtained using CLWE and QL features.

Zeta revisited: A rejoinder

Abstract Zeta is a relatively recently developed tool for authorship attribution and computational stylistics (Craig and Kinney, 2009, Shakespeare, Computers, and the Mystery of Authorship. Cambridge: Cambridge University Press). In ‘The Interpretation of Zeta Test Results’ (2019), Perves Rizvi raised several objections to the method. I responded to and rejected those objections in ‘Zeta Revisited’ (Hoover, 2021a, Zeta revisited. Digital Scholarship in the Humanities, 1–20). Rizvi has renewed two of his main objections in a critique of my response in his ‘The Interpretation of Zeta Test Results: A Supplement’ (2022). First, he reiterates and expands his argument that the separation of base and counter segments in a Zeta analysis is a mechanical consequence of Zeta, rather than a significant and meaningful finding. Second, he again rejects the use of the bisector line to attribute texts in a Zeta analysis as unsound and adds new arguments that it leads to invalid attributions. This rejoinder provides further evidence and argument that both of Rizvi’s claims must be rejected.

Audiobook Stylistics: Comparing print and audio in the bestselling segment

The paper explores differences between bestsellers in print and the most popular audiobooks in a subscription-based streaming service for books (‘beststreamers’) by means of computational stylistics. The point of departure is the complete set of print bestsellers and digital audiobook beststreamers for the Swedish book market 2015–2019, in total 172 novels. We probed 34 linguistic measures to track differences between subsets at the stylistic level. The results indicate that there are pronounced differences between the formats. Print bestsellers are longer, syntactically more complex and varied, and seem to focus more on depiction. Beststreaming audiobooks, by contrast, are shorter, more straightforwardly written, and appear to highlight plot and dialogue. The results are replicated when the comparison is restricted to crime fiction, the most prominent genre in the commercial top segment. Given these results, it is argued that it is possible to discern a particular audiobook style as one factor affecting book consumption in digital formats, and conversely that the printed format is associated with other stylistic preferences.

Analize micro-stilistice în poetica blagiană a materialității

The present study aims to highlight the main occurrences of different material elements in Lucian Blaga’s lyrical work, by focusing on quantitative and micro-stylistic approaches. The quantitative research was made with the help of a digital instrument, Sketch Engine, a relevant tool for research in corpus linguistics and computational stylistics. This study encapsulates the explicit occurences of 22 materials from Lucian Blaga’s poems, categorized into three main groups: metallic materials (ferrous and non-ferrous), non-metallic materials (ceramic/telluric) and rocks and building materials. The analysed material elements are integrated in 317 micro-structures and contexts, the main aspect observed being the ratio between metaphorical and denotative regimes in which materials are included. Starting from distant reading, the article questions a possible cognitive dimension of Lucian Blaga’s poems, which is based on the author’s cultural and philosophical experience.

Style at the Scale of the Canon. A Stylometric Analysis of 100 Romanian Novels Published between 1920 and 1940

The present study proposes an experimental exploration of the Romanian novel written between 1920 and 1940 through the use of stylometry, a method of distant reading employed for the statistical analysis of style. Drawing from the most recent advances in the field of computational stylistics, we select a formal standpoint from which we seek to investigate the relation between the Romanian novelistic canon and minor, tertiary novels published in the same. In our test cases, we will attempt to establish some of the more promising aspects of stylometric analysis, as well as single out the experiments that yield no relevant result. Because of the relative novelty of the method, the purpose of our investigations is to offer a kind of pilot experiment that can illustrate the benefits of using computational methods on Romanian literary corpora.

Big data or not enough? Zeta test reliability and the attribution of Henry VI

Abstract In 2016, the editors of the New Oxford Shakespeare announced that certain Shakespeare plays could be attributed to co-authors, and certain anonymous plays to Shakespeare, on the basis of non-traditional attribution methods known collectively as computational stylistics, or stylometry. This article investigates the efficacy of a key algorithm used to attribute parts of the Henry VI plays to Christopher Marlowe, the Zeta method invented by John Burrows and adapted by Hugh Craig. Zeta, a test widely used in computational stylistics, is described by Gabriel Egan as ‘by some way the most powerful general-purpose authorship tool currently available’. This article offers extensive independent testing of Zeta. Following criticism of the existing method of Zeta analysis, this article introduces a new, statistically sound method for analysing Zeta results. It investigates a claim that the test is 99.9% reliable in differentiating Shakespeare’s style from Marlowe’s. Examining the conditions under which certain authors were ruled in or out of co-authorship of the Henry VI plays, it determines the effect of disparity in data set size on Zeta’s reliability, as well the effect of small data sets. Several test results confirm that Zeta is unduly influenced by genre. The article concludes that in the light of this study, the small canons of most Early Modern dramatists, particularly where they are genre-skewed like Marlowe’s, do not provide enough data for Zeta to be reliable

The CLiGS Textbox: Building and Using Collections of Literary Texts in Romance Languages Encoded in TEI XML

The CLiGS textbox is published by the Computational Literary Genre Stylistics (CLiGS) group. The textbox is the group’s publication channel for several collections of literary texts. We describe the rationale for the manner in which the collections of literary texts included in the textbox have been compiled, annotated, and published. Furthermore, we suggest several ways in which the text collections can be used for research in literary studies. We aim to document some of the work of the CLiGS group, to showcase the unique TEI XML-based collections of French, Spanish, Spanish-American, and Portuguese novels and French drama we make available, and to encourage reuse of these text collections by others. We argue that agreement on common formats and procedures for text preparation, encoding, and publication fosters the accessibility, analysis, and reuse potential of literary text collections.

Style, Computers, and Early Modern Drama: Beyond Authorship. Hugh Craig and Brett Greatley-Hirsch. Cambridge: Cambridge University Press, 2017. Pp. vii+283.

<i>Style, Computers, and Early Modern Drama: Beyond Authorship</i>. Hugh Craig and Brett Greatley-Hirsch. Cambridge: Cambridge University Press, 2017. Pp. vii+283.

Language and Thought in Hildegard of Bingen's Visionary Trilogy: Close and Distant Readings of a Thinker's Development

By combining the methods of distant reading (computational stylistics) and close reading, the authors discuss the development of language and thought in Hildegard of Bingen's visionary works (Sciuias, Liber uite meritorum and Liber diuinorum operum). The visionary trilogy, although written over the course of three decades, raises the impression of a monolithic and seemingly unchanging voice. Moving beyond this impression, the interdisciplinary analysis presented here reveals that the trilogy exhibits interesting differences at the word level which cannot simply be explained through external historical circumstances (e.g. manuscript transmission or different secretaries). Instead, the results raise pertinent questions regarding the trilogy's internal development in didactic method, style, and philosophy.

(Neuro-)Cognitive poetics and computational stylistics

AbstractThis perspective paper discusses four general desiderata of current computational stylistics and (neuro-)cognitive poetics concerning the development of (a) appropriate databases/training corpora, (b) advanced qualitative-quantitative narrative analysis (Q2NA) and machine learning tools for feature extraction, (c) ecologically valid literary test materials, and (d) open-access reader-response data banks. In six explorative computational stylistics studies, it introduces a number of tools that provide QNA indices of the foregrounding potential at the sublexical, lexical, inter- and supralexical levels for poems by Shakespeare, Blake, or Dickens. These concern lexical diversity and aesthetic potential, sentiment analysis, sublexical sonority scores or phrase structure, and topics analysis. The results illustrate the complex interplay of stylistic features and the necessity for theoretical guidance and interdisciplinary cooperation in selecting adequate training corpora, QNA tools, test texts, and response measures.

Editing Aphra Behn in the Digital Age: An Interview with Gillian Wright and Alan Hogarth

This interview provides a view of the work in progress for the Cambridge University Press edition of the Complete Works of Aphra Behn. Gillian Wright serves as a general editor (with Elaine Hobby, Claire Bowditch, and Mel Evans) as well as the volume editor for Behn’s poetry. Alan Hogarth is the Postdoctoral Research Associate working with Mel Evans on the computational stylistics and author attribution testing. The discussion focuses on the scope and principles of editing the poetry of Aphra Behn, the role of stylometry in establishing the corpus, the status of work, a few particular poems, and some surprises.

Marlowe and overreaching: A misuse of stylometry

In ‘Christopher Marlowe: Hype and Hoax’(2018), Hartmut Ilsemann implies that his application of the Rolling Delta feature of R Stylo is sufficiently robust that a century and a half of traditional scholarship should be overturned, and Marlowe stripped of the majority of his canon, including Doctor Faustus and Edward II. The article concludes that ‘Marlowe is totally overrated in his influence on modern English drama’ (26), the natural consequence of stripping away 5/7ths of his canon. In this response, I demonstrate that the assumptions underlying this application of the Delta method, and the application itself, are fundamentally flawed, leading to predictably erroneous conclusions. Problems with the study include a poorly designed test environment; incorrect preparation of texts; assuming that ‘Marlowe’s style’ can be determined by a single early play; selecting and constructing Shakespeare’s comparison texts in a manner likely to prejudice results; ignoring the effect upon style of a play’s date and genre; failing to consider the effect of different-length comparison texts; dismissing external evidence of authorship that conflicts with the test outcomes. I argue that in the light of these issues, the results and conclusions must be dismissed. Further, the question is raised as to whether the current methods of computational stylistics, even when more rigorously applied, are equipped to challenge the attribution of the accepted Marlowe canon.

Hugh Craig and Brett Greatley-Hirsch, Style, Computers, and Early Modern Drama: Beyond Authorship

FOLLOWING John Burrows’ use of the commonest (function) words in English and principal component analysis as the ‘key method for what came to be known as computational stylistics’, Hugh Craig and Brett Greatley-Hirsch extend quantitative analysis to include Random Forests, Shannon Entropy, Delta, and Student’s t-test as tools of literary exploration. Random Forests and Shannon Entropy are explained in Chapter 1, although their thorough understanding requires greater familiarity than that enjoyed by literary scholars. Unfortunately for the exposition, the PCA biplot of Figure 1.3 (37) overprints Figure 1.2 (36) with scant regard for the reader. Although the focus of computational stylistics has been chiefly on authorship, Craig and Greatley-Hirsch use the extended methods of computational stylistics to engage with six questions that exclude the determination of authorship. These questions are explored individually in Chapters 2 to 7: the difference between prose and verse; the affinity (or not) of dramatic characters within and without the same authorial canon; analysis of the type and distribution of stage properties in the early modern theatre; the ‘cultural drift’ of the English language used by dramatists 1585–1624; the influence of theatre company house style on early modern dramatists in comparison with individual authorial style; the relationship of Restoration plays to plays before the Puritan closing of the theatres. The common theme is the methodology of quantitative analysis.

Quantifying the Beauty of Words: A Neurocognitive Poetics Perspective.

In this paper I would like to pave the ground for future studies in Computational Stylistics and (Neuro-)Cognitive Poetics by describing procedures for predicting the subjective beauty of words. A set of eight tentative word features is computed via Quantitative Narrative Analysis (QNA) and a novel metric for quantifying word beauty, the aesthetic potential is proposed. Application of machine learning algorithms fed with this QNA data shows that a classifier of the decision tree family excellently learns to split words into beautiful vs. ugly ones. The results shed light on surface and semantic features theoretically relevant for affective-aesthetic processes in literary reading and generate quantitative predictions for neuroaesthetic studies of verbal materials.

Bernard of Clairvaux and Nicholas of Montiéramey: Tracing the Secretarial Trail with Computational Stylistics

Bernard of Clairvaux and Nicholas of Montiéramey: Tracing the Secretarial Trail with Computational Stylistics

Mapping Illuminated Manuscripts: Applying GIS concepts to Lancelot-Grail Manuscripts

Mapping Illuminated Manuscripts: Applying GIS concepts to Lancelot-Grail Manuscripts

“We, John Cade”: Shakespeare, Marlowe, and the authorship of 4.2.33–189 2 Henry VI

In 2009, Hugh Craig acknowledged Christopher Marlowe as a probable collaborator on 2 Henry VI through the use of computational stylistics. His results most confidently identified Marlowe in 4.2.160–5.1.13, a section of the play that includes the majority of the Cade rebellion and the majority of the play’s dramatic prose. Despite the preponderance of Marlovian features in 4.2.160–5.1.13 – Craig’s data for 4.2.1–159 – a portion of the scene that includes our first exposure to Cade and the origin of the rebellion – were inconclusive. In 2011, Brian Vickers rejected Craig’s attribution to Marlowe on the basis of Craig’s failure to identify Marlowe (or any other candidate) for the authorship of 4.2.1–159. Vickers interpreted these inconclusive results as an indication that Craigrsquo;s methods were fundamentally flawed. Vickers then suggested that computational stylistics (a method that searches for the consistent preference of single words) was an inferior method when compared to traditional stylometrics (a method favored by many attribution scholars that examines word combinations). The following study seeks to evaluate how the results of stylometric testing compare to results using computational stylistics in 4.2 2 Henry VI.

Distinguishing Voices in The Waste Land using Computational Stylistics

T. S. Eliot’s poem The Waste Land is a notoriously challenging example of modernist poetry, mixing the independent viewpoints of over ten distinct characters without any clear demarcation of which voice is speaking when. In this work, we apply unsupervised techniques in computational stylistics to distinguish the particular styles of these voices, offering a computer’s perspective on longstanding debates in literary analysis. Our work includes a model for stylistic segmentation that looks for points of maximum stylistic variation, a k-means clustering model for detecting non-contiguous speech from the same voice, and a stylistic profiling approach which makes use of lexical resources built from a much larger collection of literary texts. Evaluating using an expert interpretation, we show clear progress in distinguishing the voices of The Waste Land as compared to appropriate baselines, and we also offer quantitative evidence both for and against that particular interpretation.

Computational Stylistics for Authorship Attribution - based on early Modern Korean Novels

Computational Stylistics for Authorship Attribution - based on early Modern Korean Novels

Revisiting Style, a Key Concept in Literary Studies

AbstractLanguage and literary studies have studied style for centuries, and even since the advent of ›stylistics‹ as a discipline at the beginning of the twentieth century, definitions of ›style‹ have varied heavily across time, space and fields. Today, with increasingly large collections of literary texts being made available in digital form, computational approaches to literary style are proliferating. New methods from disciplines such as corpus linguistics and computer science are being adopted and adapted in interrelated fields such as computational stylistics and corpus stylistics, and are facilitating new approaches to literary style.The relation between definitions of style in established linguistic or literary stylistics, and definitions of style in computational or corpus stylistics has not, however, been systematically assessed. This contribution aims to respond to the need to redefine style in the light of this new situation and to establish a clearer perception of both the overlap and the boundaries between ›mainstream‹ and ›computational‹ and/or ›empirical‹ literary stylistics. While stylistic studies of non-literary texts are currently flourishing, our contribution deliberately centers on those approaches relevant to ›literary stylistics‹. It concludes by proposing an operational definition of style that we hope can act as a common ground for diverse approaches to literary style, fostering transdisciplinary research.The focus of this contribution is on literary style in linguistics and literary studies (rather than in art history, musicology or fashion), on textual aspects of style (rather than production- or reception-oriented theories of style), and on a descriptive perspective (rather than a prescriptive or didactic one). Even within these limits, however, it appears necessary to build on a broad understanding of the various perspectives on style that have been adopted at different times and in different traditions. For this reason, the contribution first traces the development of the notion of style in three different traditions, those of German, Dutch and French language and literary studies. Despite the numerous links between each other, and between each of them to the British and American traditions, these three traditions each have their proper dynamics, especially with regard to the convergence and/or confrontation between mainstream and computational stylistics. For reasons of space and coherence, the contribution is limited to theoretical developments occurring since 1945.The contribution begins by briefly outlining the range of definitions of style that can be encountered across traditions today: style as revealing a higher-order aesthetic value, as the holistic ›gestalt‹ of single texts, as an expression of the individuality of an author, as an artifact presupposing choice among alternatives, as a deviation from a norm or reference, or as any formal property of a text. The contribution then traces the development of definitions of style in each of the three traditions mentioned, with the aim of giving a concise account of how, in each tradition, definitions of style have evolved over time, with special regard to the way such definitions relate to empirical, quantitative or otherwise computational approaches to style in literary texts. It will become apparent how, in each of the three traditions, foundational texts continue to influence current discussions on literary style, but also how stylistics has continuously reacted to broader developments in cultural and literary theory, and how empirical, quantitative or computational approaches have long ­existed, usually in parallel to or at the margins of mainstream stylistics. The review will also reflect the lines of discussion around style as a property of literary texts – or of any textual entity in general.The perspective on three stylistic traditions is accompanied by a more systematic perspective. The rationale is to work towards a common ground for literary scholars and linguists when talking about (literary) style, across traditions of stylistics, with respect for established definitions of style, but also in light of the digital paradigm. Here, we first show to what extent, at similar or different moments in time, the three traditions have developed comparable positions on style, and which definitions out of the range of possible definitions have been proposed or promoted by which authors in each of the three traditions.On the basis of this synthesis, we then conclude by proposing an operational definition of style that is an attempt to provide a common ground for both mainstream and computational literary stylistics. This definition is discussed in some detail in order to explain not only what is meant by each term in the definition, but also how it relates to computational analyses of style – and how this definition aims to avoid some of the pitfalls that can be perceived in earlier definitions of style. Our definition, we hope, will be put to use by a new generation of computational, quantitative, and empirical studies of style in literary texts.