Since acquiring Derrida's personal working library in 2015, Princeton University Library has embarked on a number of collaborative projects to increase access and research utility of the collection. It consists of about 13,800 published books and other materials, representing a lifetime of reading. For Derrida, the act of reading was not a passive process: he engaged—even grappled—with what he read, covering pages with notes and cross-references, inserting other handwritten materials, quoting and adapting what he read into what he wrote. As Derrida himself said in an interview later in his life, his books bear “traces of the violence of pencil strokes, exclamation points, arrows, and underlining.”

Read More about Princeton's Acquisition of Derrida's Library. https://library.princeton.edu/about/library-news/2015/princeton-university-library-acquires-jacques-derridas-personal-library

Described below are some of the past and present initiatives we've participated in alongside campus partners and collaborators.

A collaboration led by Princeton's Center for Digital Humanities, “Derrida’s Margins” is a website and online research tool for annotations from the Library of Jacques Derrida, housed at Princeton University Library. Jacques Derrida is one of the major figures of twentieth-century thought, and his library--which bears the traces of decades of close reading--represents a major intellectual archive. The first phase focused on annotations related to Derrida’s landmark 1967 work De la grammatologie (Of Grammatology). It was in Of Grammatology that Derrida first articulated a new style of critical reading, which would become the foundation of the philosophy of “deconstruction.”

Visit the Derrida's Margins Website: https://derridas-margins.princeton.edu/library/

With the success of other hand text recognition software projects such as within the Princeton Geniza Project and with the Samuel Phillips project, we have been working towards training a hand text recognition model to read Derrida’s handwriting and accelerate the transcription of materials in his archives here at Princeton to share the wealth of knowledge locked away behind his cryptic jottings.

We have opted to use Escriptorium, an open source platform for annotating, segmentation, and recognition, relying on the Kraken OCR engine as its base. Derrida’s handwriting is particularly challenging, as his idiosyncratic script is fraught with abbreviations and inconsistencies, confusing the machine just as it confuses the human reader. These challenges can be overcome with a large enough dataset of his handwriting, but even with the work of Cece Ramsey and Bill Hamlett who aligned images to transcriptions on 50 pages of Derrida’s seminars, the model is deeply imperfect and struggles to read his handwriting.

We had early hurdles in running Escriptorium on our personal machines due to the difficulties in running docker containers and managing the library dependencies for Pytorch based training, but with an institution-wide instance of Escriptorium in the works at Princeton, we hope that future researchers and completely skip over this burden and focus on the text and materials. Despite being unable to produce a high accuracy model, we hope that our early experiments will motivate scholars to consider contributing to this path of research, producing more involved TEI based transcriptions (or plain text if segmentation models improve), and helping to open up the archives of Derrida to a wider scholarly community.