At DH2015, we will be honoured to announce the creation of the new John Burrows Award. This award has been especially created for the best student or early career researcher paper presented at the next aa-DH (Australasian Association of Digital Humanities) conference in 2016. So named, its creation also acknowledges an internationally renowned pioneer in the application of computing to the analysis of style. Professor Emeritus John Burrows is the founder and remains the doyen of computational stylistics, where he has made remarkable contributions to theory and methods in forensic classification and interpretive description.
In the 1980s Burrows combined counts of very common words and Principal Components Analysis, thus founding computational stylistics. Very common words had been used before in stylometrics, notably by Frederick Mosteller and David L Wallace, but not with any idea that their patterns of use might reveal anything interesting about literary style. PCA was a well-established method in statistics but not previously been applied to language data. By 1998 Burrows’s combination of very common words and PCA had become “the standard first port-of-call for attribution problems,” in the words of David I Holmes.
Burrows’s 1987 book Computation into Criticism: A Study of Jane Austen’s Novels and an Experiment in Method combined the insights of an exceptionally gifted close reader with the new tools of computational stylistics, in a volume that is still a benchmark for the combination of interpretation and quantitative language study.
His 2001 Busa Award lecture introduced Delta, a simple and highly effective method for finding the “least unlikely” author of a mystery text among a group of candidates. The 2002 article based on this lecture is currently (late 2014) the most cited of all articles in the journal Literary and Linguistic Computing.
A 2007 article introduced two further methods, Zeta, using mid-frequency words characteristic of an author, and Iota, using rare words. This article offers evidence that authorial individuality is evident across the spectrum of word use, from the most common to the most rare. More generally, Burrows’s computational stylistics has shown that language, including literary language, is a probabilistic system, and hence (in Willard McCarty’s words) “that the most elusive of cultural qualities behaves in roughly the same way as both the social and natural worlds.”
In recent years Burrows has had a series of very fruitful collaborations with specialists who brought him attribution problems, in poetry and philosophical prose, especially, spanning the seventeenth to the twentieth century, problems he was able to resolve, with all sorts of benefits for the disciplines involved.
John Burrows will be a guest speaker during the closing ceremony, Friday 3 July.
Two ways/one way: Trad and Mod in Humanities Research
The centrepiece of this little paper is a set of stylistic comparisons among some two hundred early modern plays—184 of accepted authorship and a further thirty of uncertain or unknown authorship. This large archive of etexts has been gradually compiled and edited by Hugh Craig, assisted by Alexis Antonia.
With the five hundred most frequent words of the whole set of plays forming a base-list for frequency profiles of each play, the Delta procedure enables us to take any one play as target and to rank all the rest according to their overall level of difference from it. The process can be quickly repeatd with each play in turn as target. The effectiveness of this kind of stylistic modelling is to be assessed by the results it yields.
Elaborate as it is, all this lies well within the power of a modern laptop. Thirty years ago it would have been beyond the capacity of most mainframes. For any of our forebears it would have been a life’s work. But my purpose is not to suggest that we are cleverer than they or even to celebrate our new tools. Let us glance at some of the results.
Some of the more prominent patterns would be obvious to anyone with the least interest in these matters. Others are evident to willing amateurs like me. But the full wealth of information stored here requires the eye of experts, who will find support for many of their opinions and surprises enough to stimulate fresh ideas.
M y thesis, then, is twofold. Although it is increasingly unwise of them to do so, traditional scholars are free to ignore our offerings and confine themselves to paths pursued for many centuries. But those of us who wish to be called modern scholars rather than technicians must be well versed in the old knowledge and in the enduring values that are the glory of the humane scholarly tradition. Only so can we frame worthwhile questions and establish appropriate databases. Only so can we understand the answers. At the same time, we must continue to persuade our traditionalist colleagues that we should go forward hand in hand. While this is happening much more slowly than I used to suppose it would, we are undoubtedly gaining ground. And we shall overcome—one day.