Rachael Griffiths
Researcher
Rachael Griffiths completed her BA in the Study of Religions and History at SOAS and received her MPhil and DPhil in Tibetan and Himalayan Studies/Oriental Studies from the University of Oxford. Her doctoral research focused on the autobiographical writings of Sumpa Khenpo Yeshe Paljor (1704-1788), exploring his public identity against a backdrop of intense political change, and how these shifts impacted the intricate social, intellectual, and political networks in which he operated.
After completing her graduate studies, Rachael held a postdoctoral position in the TibSchol project at the Institute for the Cultural and Intellectual History of Asia, Austrian Academy of Sciences. She contributed to the conceptual, methodological, and technical development of digital resources, including Handwritten Text Recognition (HTR) models and a prosopographical database. Her research also traced intellectual networks and the scope of scholastic writings in the 11th-13th centuries.
She is one of the co-creators of the Oral History of Tibetan Studies project, which collects and preserves interviews with key figures in the establishment of Tibetan studies in the West. She also serves as a board member and editor-in-chief at The Digital Orientalist. This online publication focuses on the development, application, and integration of Digital Humanities in research related to Africa, the Middle East, and Asia.
Within PaganTibet…
Together with Dr Marieke Meelen, Rachael has worked on the data curation, processing and annotation side of the project. Starting with the development of a detailed Data Management Plan, then focusing on building workflows, tools, models and datasets for audio-visual fieldwork data (Automatic Speech Recognition, see Comput-EL publication on Baima), historical manuscripts (Handwritten Text Recognition, see JOHD publication), Text Classification and Information Extraction (see CHR publication) and Normalisation (see AI4CHIEF publication).
The core focus of my work so far has been developing, testing, and applying AI-assisted workflows/methods to the PaganTibet corpus. Phase one is concerned with documenting and organising the corpus - to do this we first developed iterative HTR models (using the Transkribus platform) to produce diplomatic transcriptions of our texts. As of June 2026, we have trained our final HTR model with a character error rate (CER%) of 3.71, which will be used to transcribe the remainder of the corpus.
The manuscripts within the PT corpus are challenging to read; content aside, they are full of abbreviations, spelling mistakes, and orthographic variation. To enable better and more consistent searching across and within manuscripts, we are also developing iterative Normalisation models that will standardise our transcriptions.
Links:
Academia: https://ephe.academia.edu/RachaelGriffiths
ORCID: https://orcid.org/0000-0003-1381-0718
GitHub: https://github.com/rgriffiths1
CRCAO page: https://www.crcao.fr/membre/rachael-griffiths
Cambridge page: https://www.languagesciences.cam.ac.uk/staff/rachael-griffiths