Review: Research in Humanities Computing 4 & 5

Lou Burnard
Humanities Computing Unit
University of Oxford
lou.burnard@oucs.ox.ac.uk

Selected papers from the ALLC/ACH Conferences 1992 (eds. S. Hockey & N. Ide) & 1995 (ed. Giorgio Perissinotto). Clarendon Press, Oxford 1996.

These two volumes, the latest to appear in OUP's somewhat dilatory series of publications on Research in Humanities Computing, under the general editorships of Susan Hockey and Nancy Ide, offer the reader an interesting opportunity to review activities in this field over the last five years, and thus to compare progress on a rather longer time scale than usual. The first collects 14 papers from the ACH-ALLC conference held in Oxford in 1992, while the second presents a further 15, from the ACH-ALLC conference held in Santa Barbara, some three years later in 1995. Both bear a publication date of 1996.

In 1992, the field was still dominated by literary statistics and quantitative methods. The papers by Burrows, Lancashire, Lessard and Hamm, and Opas might have been written at almost any time since the beginnings of humanities computing in the seventies: their dense tables of statistics, graphs, and diagrams bear witness to the continued vitality of a tradition that two decades of scepticism have not yet succeeded in dislodging. Whether their topic is language variation amongst writers of different nationalities (Burrows), automatic identification of phrasal collocations within a specific historic genre (Lancashire), or author (Lessard and Hamm), or multivariate statistical analysis of an author's style (Opas), these authors have all worked hard at the difficult problem of making mathematically-based reasoning accessible to a largely innumerate readership, and at clarifying their own methodological stance within an academically-respectable tradition. Whether any of the work can be said to have produced impressive new ways of reading, or provided anything more than confirmation of conclusions already reached by other methods, remain open questions.

In 1995 quantitative methods remain an important, almost an institutionalized, part of the field. Impeccable papers by Baayen, Frischer et al., and Tweedie et al. (the latter is cited as co-author in three of the papers in this volume: surely some kind of record) focus on ways of making sense out of the reams of statistical data which concordancers and the like throw up. Baayen and the Frischer team make distinct, and equally good, cases for continuing to be deeply sceptical about some of the claims of the stylometricians, while the papers from Tweedie and her collaborators continue the tradition of bending the latest new technology (here, multivariate analysis and neural nets) to the oldest problem (authorship identification).

As this last example shows, the methods and techniques of humanities computing have always been characterized by a wide ranging eclecticism. In the 1992 volume, for example, we find connectionism applied to poetic meter (Hayward); parser generator systems applied to Greek metrics (de Jong and Laan); NLP techniques developed for text understanding systems applied to reader response theory (Snelgrove); data modelling techniques applied to lexicography (Ide and Veronis) and a particularly productive application of the cladistic techniques developed in the biological and natural sciences to the analysis of manuscript variation (Robinson and O'Hara). In 1995, correspondence analysis is used to define a typology of computer mediated narrative (Aarseth), and also as a means for providing the raw data for visual presentation of a philosophical concept (Bradley and Rockwell); distributed database systems grapple with the problem of integrating text collections (Giordano, Goble and Källgren); and object-oriented systems represent complex humanistic analyses (Simons).

Perhaps the most significant preoccupation for the field, to which these volumes bear ample witness, is a focus on basic issues of text representation (or encoding) and corpus construction. Methodological papers elaborating ever more sophisticated techniques of statistical analysis remain of little use or credibility if the data analysed derive from arbitrary or academically indefensible language samples. In providing a standard method for the creators of digital resources to document and make explicit their transcription and encoding practices, the Text Encoding Initiative has laid the ground work for a revolution in the field, as well as re-centering its attention on the hermeneutic issues that characterise the humanities. It has also greatly facilitated the transfer of techniques and methods between different groups of researchers, working often in quite different disciplinary fields.

This 'cross-over effect' is particularly apparent in the way that the insights of corpus linguistics and other areas of applied linguistics have changed our notions of the enterprise of literary exegesis and analysis. The 1992 volume contains two particularly useful articles by van Halteren, on part of speech tagging, and Burr, on corpus design methods, both of which should be required reading for anyone interested in the construction of lexical corpora for whatever purpose. It also contains two nicely complementary articles on the inter-relationship between tagging and interpretation, from McCarty and Renear et al. In 1995, a year after publication of the TEI Guidelines, similar theoretical encoding issues are articulated in the context of a growing number of applications and realistically scaled projects. Examples include Hofland's practical description of an algorithm for automatic alignment of parallel corpora; Simons' description of an object-oriented implementation of the TEI's feature structure markup; and Calzolari and Monachini's account of the EAGLES proposals for standardization of morpho-syntactic analysis. The pragmatic force of such contributions is a pleasing counter-balance to the more speculative papers reviewing the limitations of the encoding enterprise, such as Neuman on manuscript transcription, or Källgren on the problems of automatically tagging Swedish.

These volumes show that humanities computing is more than a ragbag of techniques and applied statistics. It is interdisciplinary and synergistic, in the best senses of those rather overworked words, bringing insights from many fields to bear on a common set of tasks. If these volumes are to be believed, it also has a defined focus: the traditional humanistic concern with textual integrity and explication. Those embarking on an exploration of how computer-based techniques can support, hinder, or enrich that mission will find substantial guidance and information in them, beyond the simple chronicling of continuity and change within the research field they document.

[Table of Contents] [Letter to the Editor]

Computers & Texts 14 (1997), 22. Not to be republished in any form without the author's permission.

HTML Author: Michael Fraser (mike.fraser@oucs.ox.ac.uk)
Document Created: 24 May 1997
Document Modified:

The URL of this document is http://info.ox.ac.uk/ctitext/publish/comtxt/ct14/burnard.html