The Exact Analysis of Text

More than forty years after its initial publication, Frederick Mosteller and David Wallace's Inference and Disputed Authorship: The Federalist is an excellent introduction to the application of exact techniques to various problems in the analysis of text; in some respects it is still unequaled. Given the way the computer has revolutionized exact textual analysis in the last thirty years, and in particular in light of the explosion in the availability of digital texts in the last ten, it is remarkable that any work could continue to contribute for so long. When the original Addison-Wesley edition went out of print, the book was republished in 1984 by Springer (with some emendations) under the title Applied Bayesian and Classical Inference, the Case of the Federalist Papers, which, however, has also gone out of print.. We all owe thanks to CSLI publications of Stanford and publisher Dikran Karaguezian for reissuing this classic monograph in their David Hume series, specialized in mathematical studies in the Humanities, making it available once again to scholars and students. At Prof. Wallace's request, this republication bears the original, more tting title once again. Google Scholar lists several hundred citations for Inference and Disputed Authorship, most of which date since 1980; this is surely enough to demonstrate the continued impact of the book. But there are many explicit recommendations as well. In a survey article on Bayesian techniques in information retrieval in 1998 David Lewis recommended that the book by Mosteller and Wallace is the most clear treatment from a classi cation point of view (Lewis, 1998, p.7) (referring to the treatment of underlying word distributions). Mosteller and Wallace's (1964) work is also used extensively by computational linguists (Church and Mercer, 1993, Hindle and Rooth, 1993, Yarowsky, 1995), but most of all in authorship studies. Holmes and Forsyth (1995) surveys statistical authorship studies using the publication of Inference and Disputed Authorship as the watershed from which