Yesterday (Friday 3rd Aug) Melissa Terras, lecturer in Electronic Communication in the School of Library, Archive and Information Studies (UCL) gave a seminar in the Digital Classicist series at the Institute of Classical Studies entitled ‘Can Computers Ever Read Ancient Texts?’ In this excellent presentation, Dr. Terras’ answer to the question posed in the title was (pace a response on the Humanist list last week) basically ‘no’, or at least not yet and not in any significant sense as we understand “read”. A better question, she asserted, would be “Can computers assist in reading ancient texts?”, and to that question she gave (and demonstrated) a much more positive response.
In the course of this paper, which was in part a demonstration of past and ongoing work on the semi-papyrological/semi-epigraphical Vindolanda writing tablets, she made several important points that are worth repeating to an audience of epigraphists.
- The expert who teaches the computer to read texts is not principally the programmer, analyst, or engineer, but principally and essentially the papyrologist (or epigrapher) who knows the art and works in a discipline that has decades (centuries) of methodologies and expertise behind it.
- Based on Herbert Youtie’s now famous distinction between “public” (published, polished, perfect) and “private” (in progress, iterative, much-laboured) text scholarship, she points out the ability of computer-assisted readings to reveal and preserve the private processes of text editing and track methodologies and mistakes, a process known by engineers as “truth maintenance”. For example, it would be incredibly valuable to be able to document the process of making decisions as to dating and/or scribe based on letter-forms (see point 4 below), a process that seasoned papyrologists and epigraphers currently often perform based on instinct and therefore find very difficult to describe and teach.
- The computer does not, in the prefered model, make firm decisions and statements of “fact” based on engineering principles and programmed algorithms; rather it offers a range of probabilistic results (number-crunching is what machines are good at), offering likely character-groups or words, for the text editor or historian to choose between based on her expertise (interpretation is what humans are good at).
- One of the most impressive by-products of the Vindolanda text-analysis project has been the development of palaeographic markup, a system designed to record and predict average letter-forms as well as variants.
- Perhaps the most striking point was that this project, although firmly in the humanities and concerned with Ancient History, provides a rare “real-world” application of theories and techniques in Artificial Intelligence. New use-cases lead to new solutions, and a project like this benefits the discipline of Engineering Science as a whole. We Classicists should not underestimate what we have to offer to the high-powered world of medical imaging, defense technologies, and forensic science, for example.
Thanks for this summary, Gabriel… I think that pretty much sums up my stance towards the question. Although I think we need to examine what we mean by “read” – there are different ways of reading, which psychologists have known about for years, and to think about what the act of reading an ancient text actually means.
I’d hesitate to say that computers could *never* read ancient texts – who knows what technological advances are just around the corner – but in the case of the Vindolanda texts, we are hampered by what we know about the language and letter forms of the period – with every document we read, our knowledge increases – so we can only “teach” the computer to read what is known- unless some new and novel techniques are used which work in a different way to the way that humans do – my approach was to try and mimic, as best we are able, the way that experts read ancient texts, using the evidence to hand.
The system needs further development in nearly every area – I hope with the new project that palaeographic markup will come to the masses, featuring an online tool and databases….
I do firmly believe that attempting to solve difficult real world humanities problems can bring new understanding to both humanists and computing scientists/engineers. The current financial situation facing good, robust research into IT for the humanities is worrying – but thats another lecture.