Since the sophistication and wider implementation of Graphical User Interfaces (GUI) in the early 1990's, the prodigious diffusion of Hypertext has conduced a dizzying array of experiments and applications, guides and theories. Granted, many important foundations date back to prior years, decades centuries and even millennia - at least as far back as Vannever Bush's description of his Memex machine (1945) and Ted Nelson's 1965 coining of the word "hypertext" - but the more recent profusion of hypertextual technology has allowed a great pooling of resources that has combined the creative and innovative forces of many people from diverse disciplines. It's of no surprise, then, that hypertext has seen itself successfully woven into the fabric of objectives as disparate as pedagogy, retail and literary writing. Literary criticism, however, is one discipline where the hypertextual paradigm has been almost absent (TACTweb is a rare notable exception). This is all the more surprising, I think, since literary criticism stands to benefit substantially from certain possibilities offered by hypertext. Time and again colleagues have commented on the regrettable chasm that continues to exist, methodologically, between the text to be analysed and the data produced to help analyse it (retrospectives on computing in the humanities such as those by Raben and Potter invariably invoke this impediment). This paper will present the latest version of HyperPo (Hypertexte potentiel), a Web-based application intended, among other things, to bridge the gap between text and data by fully exploiting the hypertext paradigm (HyperPo is available at http://qsilver.queensu.ca/QI/HyperPo).
It needs to be asserted from the outset that HyperPo is as much a tool for text exploration as a program for text analysis (a fundamental characteristic enshrined in its full name "HyperPo: Text Analysis and Exploration Software"). In fact, HyperPo strives to blur the distinction between exploration and analysis with two related strategies: (1) the superimposition of text and data through typographical schemes (2) the tight integration of text and data through hypertextual links. The act of reading the text and the acts of producing and analysing data are all collapsed into a flexible exploratory and experimental activity that assumes the modalities desired by the particular reader/user/researcher (these distinctions are of course also blurred).
In essence, HyperPo is an extensible text reading program. The first-time user can have HyperPo display an electronic text with little or no supplementary information (in a format that would seem comforting and familiar), and progressively add types of textual information deemed potentially interesting or useful. I see this incremental approach to computer-assisted literary criticism as a crucial step if it is to gain wider acceptance. No longer must the traditional literary critic make a daunting leap of faith away from her text and towards a new world of quantitative analytical tools; she can stroll around on relatively familiar ground and, in time, explore and claim new territories as she sees fit. Although HyperPo's user-friendliness doesn't come at the expense of utility and power, it should be conceded that all of its text analysis functions (frequency, distribution, cooccurrence, KWIC lists, etc.) are available from other programs. What is novel, and epistemologically significant (in ways that will be defined and elaborated), is the possibility of dynamically generating text and data intertwined in a rich web of associated information.
Given that HyperPo is freely available as a web-based program -- which, among other advantages, eliminates the need for the user to install and update software other than a browser -- I will forego a detailed description of the software in favour of a brief account of its major features and a somewhat hypothetical presentation of it in action with an excerpt of Georges Perec's La Disparition. (My work on the play between literary game and violence in La Disparition will be a secondary preoccupation in this context. Although I heed the warnings of colleagues to not concentrate on the tools of criticism at the expense of proper literary scholarship, I think that HyperPo represents a sufficiently significant shift in methodology that the process of hypertextually driven exploration merits full consideration and elaboration. See my ACH-ALLC'97 summary for an apologia of Oulipian texts in my corpus).
As one would expect, the absence of the letter 'e' in Perec's lipogrammatical novel La Disparition severely limits the use of the primary feminine morphological marker in nouns and adjectives of French. This in turn further limits an already deleted vocabulary, encouraging recourse to foreign and unusual words («la Star» and «la Squaw»), unorthodox paraphrases («pli labial» for une lèvre) and aberrant syntactical forms («un Roman Jakobson qui nous dirait son structural avis» -- the strange anteposition of the adjectif masks how Perec avoids more common forms such as «structuralisme» or structuraliste»). In fact, the lipogrammatical constraint propagates itself through successive textual levels: the graphemic, the morphemic, the lexical, the syntactical, the textual and the semantic (the precise meaning and order of these are obviously open to discussion).
By representing parts of speech or repetition information directly in the text (through a chosen colour scheme), by linking together text based on this information, by providing the means to generate KWIC, distribution and coocurence lists, and finally, by linking words to outside resources such as the Altavista search engine and the ARTFL Frequency database, HyperPo provides the means of exploring different stylistic strategies present in La Disparition and indeed in any text. Figure 1 shows a screen shot of HyperPo where the relative frequency of repeating series of two words are shown in increasing hues of grey in the text of the top left frame, the list of repeating series of words in alphabetic order are shown in the top right frame, the distribution of pouvoir is shown in the bottom left frame, and other information is indicated in the status bar.
As always, HyperPo is about making readily available a relatively powerful but easy-to-use tool for text exploration and analysis; one that fully exploits the hypertextual paradigm. Despite progress, much frustration has resulted in Humanities Computing from trying to make computers do things that they can't (yet?) do. HyperPo recognizes the respective strengths of humans and computers and attempts to maximise the collaboration through hypertext.