This talk presents the principal research questions, methods, and some preliminary results of a rather young project, MASLOTS—Mapping Scribal Landscapes of Tibetan and Sanskrit. MASLOTS utilises deep learning (DL) assisted scribe identification in the study of two manuscript groups from different historical periods that were both crucial for the spread of Buddhism across Asia. We first provide a general outline of the DL setup, which builds upon and extends Peer, Kleber, and Sablatnig 2023; Peer and Sablatnig 2023. We then describe the format and extent of the data used as ground truth for this setup. Finally, we discuss some initial findings and evaluate the utility of the DL methods for the scholarly editing of a text from the second group of manuscripts, which is linked to Indian scholars who travelled to Tibet at the beginning of the 13th cent. CE.

To understand the presentation’s aims, it is important to know some of the historical background to the texts that are being examined. Upon arriving in Tibet from east India in 1204 with a retinue of students and junior colleagues, the Kashmiri “great scholar” (mahāpaṇḍita) Śākyaśrībhadra (Śākyaśrī for short, ca. 1127–1225) seamlessly continued his already illustrious career (Stearns 1996). His and his followers’ endeavours in early 13th century Tibet contributed to a pivotal moment in the Tibetan assimilation of Indian culture in two areas: Buddhist esoteric literature and practices centring on the Kālacakra-tantra, and the literature and ideas of the tradition of Buddhist logic and epistemology (pramāṇa). Indeed, Śākyaśrī and several of his followers were not only accomplished philosophers and logicians, but also tantric adepts specialized in the Kālacakra system. Although most of the works that Śākyaśrī and his followers wrote or translated are tantric, at least two of the disciples, Vibhūticandra and Dānaśīla, are closely associated with Sanskrit manuscripts of philosophical texts. Both these scholars can be traced through manuscripts that survive to the present day, notwithstanding the different careers they had after their arrival in Tibet.

Tibetan works retain many details of Vibhūticandra’s life due to his significance in Buddhist traditions in Tibet (Stearns 1996). He refers to himself as the “east Indian paṇḍita Vibhūticandra,” having been trained in both worldly and Buddhist (sūtra, vinaya, abhidharma) fields of learning in eastern India, while also receiving initiations in the Kālacakra-tantra (Scheuermann 2019). Once in Tibet, Vibhūticandra contributed significantly to the dissemination of the “sixfold yoga” (ṣaḍaṅgayoga) teachings from the traditions of Śavaripa (12th–13th cent.) and Anupamarakṣita (10th–11th cent., see Stearns 1996:  136–146 for both).

Dānaśīla was also part of Śākyaśrī’s retinue. He was regarded as an authority in the field of Indian Buddhist epistemology, specialising in the works of two major Indian Buddhist philosophers, Dharmottara and Prajñākaragupta (see Van der Kuijp 2003:  410). Even though his influence in Tibet is not comparable to that of Vibhūticandra, he did teach at least two famous Tibetan scholars in the subjects of Buddhist epistemology and logic (Stearns 1996:  133; Van der Kuijp 2003:  410): the influential Sa skya scholar and politician Sa skya paṇḍita Kun dga’ rgyal mtshan (Sa paṇ), and bCom ldan ral gri (or Dar ma rgyal mtshan, 1227-1305),

Surviving manuscript material reveals the considerable extent of Vibhūticandra’s and Dānaśīla’s interest in Buddhist epistemology and logic. Vibhūticandra composed extensive glosses and an appendix (kroḍapattra) to the Pramāṇavārttikavṛtti, Manorathanandin’s commentary on Dharmakīrti’s Pramāṇavārttika (Watanabe 1998a:  iv–v). This appendix contains multiple pages of notes on the Pramāṇavārttika. The precise purpose of the glosses and the extended comments in the appendix, however, remains unclear.

This presentation will focus on the scholarly edition of Vibhūticandra’s glosses pertaining to the third chapter, the “perception” (pratyakṣa) chapter, of Manorathanandin’s Pramāṇavārttikavṛtti (Sāṅkṛtyāyana 1938–1940) as well as of the commentary added to the end of the manuscript of that work (PVVms).

The edition of this material has to fulfil two functions: it must allow the extraction of data suitable for training scribal identification algorithms (essentially linking the transcribed text to image data), and it must be convertible into a printed edition suitable for publication. To achieve this, the annotated edition adheres to a customised set of the TEI Guidelines (TEI P5 Guidelines) that is under development under the name of “Vienna Encoding Guidelines for Editing Sanskrit Texts” (VEGEST schemas). The raw XML edition itself constitutes a component of the research data generated by MASLOTS, and can be archived in full accordance to FAIR principles (Wilkinson et al. 2016).

A question to discuss is whether the identification of other manuscripts in the scribe’s hand (or at least of manuscripts that are close to that hand) facilitates the quality of the edition. One common problem in editing badly preserved material in abugida writing systems is that certain combinations of characters (akṣara-s in many South Asian scripts) are quite rare, and a low number of occurrences can result in uncertain decipherment of imperfectly executed or badly preserved instances. A second obstacle are characters that are only partially legible. This obstacle is especially serious for material found in the margins of manuscripts, where the handling of the leaves has the most destructive impact. How many of these problems can be resolved once a significant amount of similar hands has been identified?

A second and wider question concerns the extent to which an author’s intention should be considered in editing texts of this genre. Typically, for Indian Buddhist philosophical works the lack of details about authors’ lives makes it impossible to answer this question. However, Vibhūticandra’s case might differ. For example, we will consider whether Vibhūticandra’s glosses and notes reflect Śākyaśrī’s aim to establish what he considered an improved and corrected interpretation of foundational texts in Indian Buddhist philosophy. Since Vibhūticandra’s works are characterised by a very high level of intertextuality, the edition will benefit from the significantly greater availability of parallel materials compared to the mid-20th century when Sāṅkṛtyāyana undertook his editions (Frauwallner 1960; Steinkellner 1981).