Academia.eduAcademia.edu

The Library of Nonhuman Books

2019, Proceedings of xCoAx International conference on Computation, Communication, Aesthetics & X

The Library of Nonhuman Books centres around a custom-made reading-machine which uses machine-learning to abridge and ‘artificially illuminate’ physical books through a combination of algorithmic interpretation and digital palimpsest. Newly illuminated texts are offered as alternative futures of the book. The project speculates on the book to come, where a post-literate society defers its reading to nonhuman counterparts.

xCoAx 2019 Karen ann Donnachie [email protected] Independent Artist, Melbourne, Australia Andy Simionato [email protected] Conference on Computation, Communication, Aesthetics & X Milan, Italy The Library of Nonhuman Books RMIT University, Melbourne, Australia Keywords nonhuman reading natural language processing illumination generative writing installation The Library of Nonhuman Books centres around a custom-made reading-machine which uses machine-learning to abridge and ‘artificially illuminate’ physical books through a combination of algorithmic interpretation and digital palimpsest. Newly illuminated texts are offered as alternative futures of the book. The project speculates on the book to come, where a postliterate society defers its reading to nonhuman counterparts. 1 THE PROJECT Fig. 1. The Library of Nonhuman Books [reading machine], 2019. The Library of Nonhuman Books is an autonomous reading system, intended to function as a “bridge” between human and nonhuman readers of books produced between the mid-15th and late-20th century. At the heart of this system is a custom-made reading-machine which can interpret any page of text from a (physical) book placed in its ‘cradle.’ The machine algorithmically searches for combinations of words on the page in an attempt to create new derivations of the original text. These combinatorial results are revealed through the erasure of all other words on the page. Based on the resulting text, the system then searches online for an accompanying image that it uses to ‘illuminate’ that page. The resulting spread is projected in real time into the exhibition space, allowing the reader to contemporaneously view both the original book (which lies within the machine) and its altered pages (projected onto the wall or screen). Automation of this system also permits the creation of new editions, which are occasionally printed, bound, and added to the Library of Nonhuman Books (Fig. 2). In addition to the reading-machine, a small version of this Library is exhibited. Fig. 2. The Library of Nonhuman Books [Sample of printed Artificially Illuminated book], 2019 1.1 An Artificially Illuminated script. This project creates a digital palimpsest, or scraping away, of texts, which are offered in the place of, or alongside, the original material book. Working within the genre of erasure poetry with the added disruption of ‘illumination,’ our reading-machine leverages artificial intelligence and machine learning to make and unmake meaning with every turn of the page (Figs. 3-5). Significantly, the system attempts to perform its combinatorial functions while preserving the materiality of the book-object, including the physical texture of the book’s support, and other para-textual elements such as page numbers and marginalia. When a book is placed under the reading-machine, the system automatically and continuously generates new combinations of the printed text, and adds semantically related images taken from the internet. Through this process of ‘artificial illumination’ new meanings are proposed that, until that moment, have remained latent in the original work, not consciously perceived by the reader. Fig. 3. The Library of Nonhuman Books [spread]: “The Handmaid’s Tale” (Illuminated), 2019. Spread from the algorithmically modified edition of Margaret Atwood’s ‘The Handmaid’s Tale’ (1995). Fig. 4. The Library of Nonhuman Books [spread]: “The Terminal Beach” (Illuminated), 2019. Spread from the algorithmically modified edition of J G Ballard’s ‘The Terminal Beach’ (1966). Fig. 5. The Library of Nonhuman Books [spread]: “The Psychology of Perception” (Illuminated), 2019. Spread from the algorithmically modified edition of Vernon’s ‘The Psychology of Perception’ (1962). 2 TECHNICAL DESCRIPTION Fig. 6. The Library of Nonhuman Books [installation diagram]. Reading machine, books and digital projection, 2019. This Nonhuman Reading system is custom designed, and comprises a camera, a Raspberry Pi and a projector. The machine’s camera ‘reads’ the book placed in its field of view, by way of an original Python script leveraging Machine learning and AI, with OpenCV and NLTK (Python Natural Language ToolKit). The algorithm parses, interprets, modulates and transforms the text by removing a substantial number of the words to leave a grammatically determined ‘essence’ of the original text in its original position on the page. The reading-machine’s default setting searches for a low-density syllable-based new reading for deriving semantic meanings, reminiscent of a Haiku poem. However, alternative scripted variations of our Natural Language Processing (NLP) -based system can be activated to further inter- pret the printed word through semantic synthesis (eg. abridgement) and/ or lexical visualisation (eg. concrete poetry). The remaining work is then ‘illuminated’ through the addition of a semantically associated line drawing sourced from Creative Commons Google Image search results, and projected in real time. Additionally, a selection of ‘pre-read’ books have been printed and are offered for perusal. Fig. 7. The Library of Nonhuman Books, 2019.https://vimeo.com/317774418 3 ACKNOWLEDGMENTS At the heart of this project, our custom-made human-machine reading system runs in Python on Raspberry Pi (Debian) and leverages open source libraries—Computer Vision (CV), Optical Character Recognition (OCR), Natural Language Processing (NLP), and Google Image API. We would like to acknowledge our support and appreciation of the authors and contributors of these projects. Additionally, Andy Simionato recognises RMIT University for research travel assistance. All media materials in this paper courtesy Karen ann Donnachie and Andy Simionato (Donnachie, Simionato & Sons), 2019.