The Spectator press is a pan-European periodical genre of the 18th century renowned for the deliberate play of authorial disguise: instead of writing under their real names, the spectatorial authors introduce a fictitious author-character. In the most prominent model of these periodicals – The Spectator (1711-1712, 1714) – the real authors, Joseph Addison and Richard Steele, create the fictitious author-character of ‘Mr. Spectator’ who, as an anonymous and confidence inspiring observer, communicated with the readers by exchanging letters with them. Yet, not only Addison and Steele but also other collaborators hid behind the eidolon of ‘Mr. Spectator’. However, the real authors tagged the individual issues by a code of letters, which enables the readers of the time and contemporary literary scholars to attribute issues of respective periodicals to a specific author (cf. Rau 1980, 74).
Throughout the 18th century, all Spectator emulations adopted the mask of the fictitious author, but largely lacked these authorial markers, making it difficult to attribute them to a specific author with close reading literary analysis methods. To counteract this problem, distant reading methods, such as stylometry, can provide significant insights into the text corpus of the pan-European Spectator press, especially regarding the attribution of presumably anonymous and collaborative authorship.
The publicly available digital edition of the European Spectators, implemented at the University of Graz, contains more than 3.500 single issues of periodicals in Spanish (647 issues), French (1672 issues), and Italian (1320 issues) which are encoded using the standard of the Text Encoding Initiative. Many of the periodicals are published anonymously or by a group of authors. Based on this corpus, the Distant Spectators project, a cooperation of the Centre for Information Modelling, Institute of Romance Studies (both University of Graz), the Institute of Interactive Systems and Data Science (Graz University of Technology) and the Know-Center Graz, examines the application of various distant reading methods to a multilingual literary corpus and cross-references the results with a well established close reading literary research expertise. Our methods and results developed in the course of this project will also be open sourced, enabling other researchers in the field to conduct similar studies.
In this contribution, we discuss how the application of specific stylometric methods can be employed to infer the identities of anonymous authors or translators, while bearing in mind the limitations of only a few comparable texts of the individual authors and the relative brevity of single issues of the periodicals. Specifically, we start by computing Burrow’s Delta as a first baseline approach. Our features include tuples consisting of words and their respective part-of-speech tags. We then use all issues of Spectator periodicals for which the authorship is clearly known and create hold-out test sets for each of them (depending on the number of issues written by respective authors with a maximum test set size of 10 issues per author). Our results are promising, achieving a mean performance of 81% across all three languages, meaning that, on average, we can detect the correct author for 81% of issues written in a respective language. This suggests that such methods are applicable to the Spectator periodicals.
An outstanding example of unsolved anonymous redaction is the French translation of The Spectator, Le Spectateur ou le Socrate moderne, which was published anonymously between 1714 and 1726. While Martens (1968, 25) suggested a possible participation of Van Effen in the translation of the text, Rau (1980, 153-159) and Lévrier (2007, 42-44) considered it unlikely. To investigate this case through stylometric methods, we first compute similarities between texts using the Stylometry with R package based on term frequencies of all French periodicals from known authors and Le Spectateur ou le Socrate moderne, representing the result as a bootstrap consensus network (Eder 2017). Here, the translations are close to works of Van Effen and La Chapelle (cf. Figure 1), partly confirming the theory of Martens (1968, 25).
Figure 1: Bootstrap Consensus Network of French Spectator periodicals
Due to the fact that several authors published translations as well as their own writings, in this analysis we represent distinct periodicals of all authors as separate nodes. Note that the grouping of all three publications of Van Effen, but also those of Bastide, Marivaux and Delacroix indicate that the individual style of the authors is detectable through the network. Also, we note the close clustering of Le Mentor Moderne (translation of The Guardian by Van Effen), Le Philosophe nouvelliste (translation of The Tatler by Boisbeleau La Chapelle) and the anonymously published Le Spectateur ou le Socrate moderne. This observation could indicate that the style and choice of vocabulary in this example was determined by the style of the original texts, all originating from the English authors Joseph Addison and Richard Steele, as suggested by Rybicki (2012).
To verify these results and further improve authorship detection we employ the General Imposters method comparing the Le Spectateur ou le Socrate moderne with texts from possible authors. Using different distance measures and most frequent words samples revealed Van Effen as the most likely author with certainty values between 0.82 and 0.95 which corroborates Martens claim.Bibliography:
Addison, Joseph/Richard Steele (1711-1714): The Spectator. Edition by Donald F. Bond
(1965), 5 volumes, Oxford: University Press.
Anonym (1716-1726): Le Spectateur, ou le Socrate moderne, Où l'on voit un Portrait naïf des Mœurs de ce Siècle. Traduit de L'Anglois. Paris: Etienne Papillon/Frères Wetstein/François-Guillaume L'Hermite. Edition by Ertler/Fischer (2011-2019).
Eder, M. (2018). Authorship verification with the package stylo. https://computationalstylistics.github.io/docs/imposters
Eder, M. (2017). “Visualization in stylometry: cluster analysis using networks.” Digital Scholarship in the Humanities, 32(1): 50-64
Eder, M., Rybicki, J. and Kestemont, M. (2016). “Stylometry with R: a package for computational text analysis.” R Journal, 8(1): 107-21. https://journal.r-project.org/archive/2016/RJ-2016-007/index.html
Ertler, K.; Fuchs, A.; Fischer, M.; Hobisch, E.; Scholger, M.; Völkl, Y. (eds.) (2011-2019): The Spectators in the International Context, https://gams.uni-graz.at/spectators.
Lévrier, A. (2007): Les journaux de Marivaux et le monde des « spectateurs ».
Paris: Presses de l’Université de Paris-Sorbonne.
Martens, W. (1968): Die Botschaft der Tugend. Die Aufklärung im Spiegel
der deutschen Moralischen Wochenschriften. Stuttgart: Metzler.
Rau, Fritz (1980): Zur Verbreitung und Nachahmung des 'Tatler' und 'Spectator'. Heidelberg: Winter.
Rybicki, Jan (2012): “The great mystery of the (almost) invisible translator. Stylometry in translation”, Quantitative Methods in Corpus-Based Translation Studies, edited by Michael P. Oakes and Meng Ji. John Benjamins Publishing Co, 231-248.