The GI therefore proposes the following iterative procedure, which can be likened onesto forms of ‘bootstrapping’

The GI therefore proposes the following iterative procedure, which can be likened onesto forms of ‘bootstrapping’

Let x represent an unknown document and let y represent per random target author’s stylistic ‘profile’. During one hundred iterations, it will randomly select (a) fifty per cent of the available stylistic features available (anche.g. word frequencies) and (b) thirty distractor authors, or ‘impostors’ from per pool of similar texts. Sopra each iteration, the GI will compute whether interrogativo is closer sicuro y than preciso any of the profiles by the thirty impostors, given the random selection of stylistic features durante that iteration. Instead of basing the verification of the direct (first-order) distance between incognita and y, the GI proposes esatto record the proportion of iterations durante which interrogativo was indeed closer sicuro y than preciso one of the distractors sampled. This proportion can be considered a second-order metric and will automatically be a probability between nulla and one, indicating the robustness of the identification of the authors of incognita and y. Our previous work has already demonstrated that the GI system produces excellent verification results for classical Latin prose.31 31 Compare the setup per Stover, et al, ‘Computational authorship verification method’ (n. 27, above). Our verification code is publicly available from the following repository: This code is described per: M. Kestemont et al. ‘Authenticating the writings’ (n. 29, above).

For modern documents, Koppel and Winter were even able sicuro report encouraging scores for document sizes as small as 500 words

We have applied a generic implementation of the GI sicuro the HA as follows: we split the individual lives into consecutive samples of 1000 words (i.di nuovo. space-free strings of alphabetic characters), after removing all punctuation.32 32 Previous research (see the publications mentioned con the previous two notes) suggests that 1,000 words is verso reasonable document size durante this context. Each of these samples was analysed individually by pairing it with the profile of one of the HA’s six alleged authors, including the profile consisting of the rest of the samples from its own text. We represented the sample (the ‘anonymous’ document) by verso vector comprising the incomplete frequencies of the 10,000 most frequent tokens sopra the entire HA. For each author’s profile, we did the same, although the profile’s vector comprises come per vedere chi si ama sul meetmindful senza pagare the average divisee frequency of the 10,000 words. Thus, the profiles would be the so-called ‘mean centroid’ of all individual document vectors for a particular author (excluding, of course, the current anonymous document).33 33 Koppel and Seidman, ‘Automatically identifying’ (n. 30, above). Note that the use of a single centroid verso author aims to veterano, at least partially, the skewed nature of our data, since some authors are much more strongly represented mediante the corpus or preparazione pool than others. If we were not using centroids but mere text segments, they would have been automaticallysampled more frequently than others during the imposter bootstrapping.

To the left, a clustering has been added on vertice of the rows, reflecting which groups of samples behave similarly

Next, we ran the verification approach. During one hundred iterations, we would randomly select 5,000 of the available word frequencies. We would also randomly sample thirty impostors from per large ‘impostor pool’ of documents by Latin authors, including historical writers such as Suetonius and Livy.34 34 See Appendix 2 for the authors sampled. The pool of impostor texts can be inspected per the code repository for this paper. In each iteration, we would check whether the anonymous document was closer to the current author’s profile than esatto any of the impostors sampled. In this study, we use the ‘minmax’ metric, which was recently introduced per the context of the GI framework.35 35 See Koppel and Winter, ‘Determining if two documents’ (n. 26, above). For each combination of an anonymous text and one of the six target authors’ profiles, we would superiorita the proportion of iterations (i.e. verso probability between niente and one) durante which the anonymous document would indeed be attributed esatto the target author. The resulting probability table is given mediante full con the appendix puro this paper. Although we present a more detailed discussion of this data below, we have added Figure 1 below as an intuitive visualization of the overall results of this approach. This is a heatmap visualisation of the result of the GI algorithm for 1,000 word samples from the lives sopra the HA. Cell values (darker colours mean higher values) represent the probability of each sample being attributed esatto one of the alleged HA authors, rather than an imposter from a random selection of distractors.