Reading Machines: Toward an Algorithmic Criticism (Topics in the Digital Humanities) [Paperback]

The first thing that can be said about this book is that in spite of the title it is not one that discusses how to train a machine to perform textual criticism. Therefore it may disappoint those readers who made that assumption about the content of the book. That being said, the discussions in the book may bring the development of such a machine to fruition, and no doubt the author would appreciate the completion of such a project. He is cautious about the use of scientific inquiry in textual criticism, as brought out in the introduction to the book, and he makes it clear that such an endeavor is not the purpose of the book, but one can argue with some justification that `algorithmic criticism' of the type that could be implemented in an intelligent machine would be in line with what he proposes in the book. Such a machine would be able to "gain entry to the deliberately and self-consciously subjective act of critical interpretation" to paraphrase the author, and such an ability he states is not yet available.

Hence the `algorithmic criticism' as part of the `digital humanities' that the author considers a legitimate method of inquiry is discussed in a different context in this book. All of the discussion is interesting, and will in fact be refreshing for those readers who are fed up or irritated by "deconstructionist" methods of textual criticism that have been popular for the last few decades. It might appear sometimes that the author is making an intersection with these methods, such as when he makes statements about the "estrangement and defamiliarization of textuality", but these are made only for contrast, and the author quickly makes the transition to more sensible and optimistic discussions of algorithmic textual analysis. A particular source of fascination is the discussion in the book about a theory of a whole collection of novelistic forms, since this would deliver a classification of writing that enabled one to understand to what degree a particular novel is "different" than another one. This type of understanding would occur in a kind of "conceptual space" and not a purely syntactical one, and would find a possible home in current studies on the cognitive neuroscience of reading. Brief reference is also made to the use of principal component analysis and statistical clustering methods to compare texts. Readers with experience in the use of these methods may be motivated to investigate in more detail their use in comparing for example Romantic and Renaissance tragedy as was mentioned in the book (but unfortunately not discussed in detail).

Scientists of course do not need to be reminded of the enormous role played by the creative imagination in scientific discovery, and therefore it could be said that the author is devoting too much time to discussion on the purported schism between the humanities and the sciences regarding this role. Yes, scientific methodologies employ quantitative methods, but the scientific discovery process has resisted quantitative characterization and formalization so far. Those who are attempting to automate the scientific discovery process using intelligent machines understand this very well. Hence the discussion on "thought experiments" and "pataphysics" in the book, which is characterized as a "science of imaginary solutions" seems somewhat vacuous if it were not for the author's discussion on how science could get bogged down in "common consent and habit" if the imagination were not relatively unconstrained in the discovery process. Even further, the author points to "imaginative meaning" as an interplay between discovering what is possible while still respecting some sort of constraints. Scientists do this on a daily basis, and the author wants to use this as incentive and inspiration to do the same in the construction of literary works.

So finding the possibilities of literary works, coming under the topic of "potential literature" follows a similar path, namely not total anarchy but one that is still fully aware of the combinatorial explosion of literary forms. The literary works that are eventually constructed will reflect the choice of solutions that are all considered equally probable, but that still respect the "rigid constraint of form." As examples, the author refers to poetry that could be generated by algorithmic processes, and the use of algorithms to analyze hidden meanings/interpretations. This also includes "potential readings" of a text, both in and out of context. The author's reference to the possible permutations and realizations of texts has similarity to that of `ergodic' literature. The author devotes a small amount of space in the book to this interesting topic, and such an inclusion is justified. Indeed, the key strategy of ergodic literature is to require the reader to be actively involved or interact with the work at hand. One popular example (other than hypertext), and one in which the author discusses, is the ancient Chinese text I Ching. The Book of Changes uses randomization to combine the texts of the `hexagrams.'