Last October 2024, I had the great privilege to be invited to give a paper at “The International Conference on Turfan Study: New Developments on the Research of Jingjiao” in the city of Turpan (Central Asia – West China), where I was honored to present alongside leading scholars working in the field of Turfan Studies from around the world. The conference was held within the Turpan Museum. During a coffee break, a kind Chinese student took me to see newly discovered materials in the museum, where my attention was drawn to a unique Syriac fragment attached to other fragments with Chinese script.
The lighting was dim, the script faded, and distance prevented close inspection. After reading the visible Syriac words a couple of times, the familiar phrases resonated within me. In that moment, relying on nothing but my memory of Syriac prayers, a skill no algorithm can yet replicate, I recognized the familiar cadence of Psalm 91, and I started to recite the Psalm in the museum: Yes, it was from the Sutoro compline prayer “ܝܬܒ ܒܣܬܪܗ ܕܡܪܝܡܐ”!
This experience, far removed from the digital datasets that fuel Artificial Intelligence (AI), underscores a crucial point: while AI tools like Handwritten Text Recognition (HTR) have made remarkable steps in automating transcription (as discussed in my previous post on Syriac HTR on Transkribus, here), they still face significant hurdles. AI algorithms struggle with the realities of manuscript research: faded ink, torn fragments, inconsistent scripts, and the slight signs that reveal liturgical function. In such cases, human expertise—the ability to connect textual fragments with a broader historical and liturgical context—becomes indispensable.
My fascinating experience of identifying the fragment at Turpan museum also resonates with an earlier experience I had of unexpectedly finding significant Syriac fragments housed at the Beinecke Rare Book & Manuscript Library, Yale University (collection: Hartford Theological Seminary), where, as it often happens, many valuable texts “flew under the digital and non-digital scholarly radars” (see part 1 and part 2). My extraordinary finding at Yale was an important lesson, to realize that not everything can be found in the digital or digitized catalogues, but in many cases, it is the human zeal and knowledge that can lead to incredible findings in the world of manuscripts.
The Syriac-Chinese Turfan Fragment: A Glimpse of Psalm 91
The above-mentioned fragment in the Turpan Museum (as far as I could observe from behind its glass case) is made of paper inscribed with Syriac text. It appears to be a remnant of a small manuscript that was bound together with recycled fragments, including one bearing Chinese script. The visible Syriac words confirmed it as a portion of Psalm 91: 7c–13, according to the Syriac Peshitta.
Here is the complete Syriac text of the fragment that I have constructed from what I could see (the black color refers to what could be read, and the red color represents the missing text in the fragment):
The fragment’s liturgical context is significant: it aligns with the East Syriac morning prayer for Sundays and feasts (in Qdam W-Bathar), and the West Syriac compline prayer (Sutoro) traditions.
Syriac Liturgical Presence Along the Silk Road
The presence of Syriac liturgical texts in Turfan is not an isolated occurrence. From around the 8th century onwards, Syriac Christian merchants and monks traveled from Mesopotamia along the Silk Road to central Asia and China, establishing monasteries in places like Turfan. Those monks carried their liturgical traditions, including a rich collection of Syriac texts. Thanks to the remarkably dry weather of Turfan, the fragments were preserved in the sand for many centuries. The Syriac fragments found in the region contain prayers shared by both East and West Syriac traditions, highlighting Turfan’s role as a meeting point of diverse liturgical practices.
The Role of Human Intelligence
The human ability highlighted in the above story relied on more than just recognizing individual letters. If I think about it now, my capability to reach such a result required:
- Linguistic knowledge: A proper understanding of Syriac script variations, allowing me to decipher faded and incomplete words (so of course, you need first to learn Syriac!)
- Liturgical memory: The ability to recall Syriac prayers and liturgical formulas, providing a context for interpreting the fragment’s content (try to attend Syriac prayers?)
- Contextual understanding: Familiarity with the historical development of Syriac liturgy, enabling me to place the fragment within a broader tradition.
Current AI tools for Syriac—and likely for other languages as well—are powerful, but they often struggle with fragmentary, damaged, or contextually ambiguous manuscripts. Human scholars bring a nuanced understanding that remains essential for accurate interpretation. Today, we often encounter overly optimistic projections about how AI might function in the real world—echoing the early days of the computing revolution, when imagination sometimes outpaced reality. From my experience working extensively with both manuscripts and digital technology, I have found that human insight, informed by knowledge and experience, remains crucial for advancing the field of manuscript studies. It is this understanding that guides my current project at the Austrian Academy of Sciences.
The “Identifying Scattered Puzzles of Syriac Liturgy” (ISP) Project and Turfan Fragments
My encounter with the Turfan fragment aligns directly with the objectives of the ongoing FWF project: “Identifying Scattered Puzzles of Syriac Liturgy” (ISP), based at the Austrian Academy of Sciences in Vienna. As discussed in my earlier post, ISP seeks to build a Syriac Liturgical Corpus by analyzing scattered fragments and tracing the historical diffusion of liturgical texts. The project’s aims include identifying textual matches, reconstructing missing portions, and providing liturgical and historical context for these fragments. Its progress is publicly accessible through the project’s website: https://syriac-liturgy.org.
While ISP integrates several digital tools into its workflow—including AI-powered Handwritten Text Recognition (HTR) models via Transkribus, dataset management through GitHub, and the use of TEI files for encoding transcriptions—human scholarly expertise remains essential. It is ultimately through careful philological analysis and contextual knowledge that accurate identifications of fragments and manuscripts are made. Digital tools enhance our capabilities, but it is human intelligence that guides interpretation and meaning.
Concluding Words
The Syriac fragment housed in the Turpan Museum offers a rare and meaningful glimpse into the vibrant liturgical life of Syriac Christianity along the Silk Road. Its identification—achieved through human memory and scholarly expertise—highlights the importance of traditional philological skills in our digital age.
For those looking to AI to solve all their manuscript problems, this case serves as a timely reminder: while artificial intelligence holds immense promise for accelerating manuscript research, it cannot replace the depth of insight and intuitive connections that human scholars bring. The future depends on a collaborative model, where AI supports rather than supplants scholarly discernment. As we continue refining these technologies, let us not forget the unique interpretive capabilities of the human mind.
Just as crucial is the engagement with heritage scholars—those deeply familiar with living Syriac liturgical traditions. Their insight is essential to identify and describe texts that remain in active use. To fully appreciate the significance of this Turfan fragment, I encourage interested readers to attend a Ramsho (evening prayer) service in a Syriac church, where the liturgy traditionally concludes with the Sutoro prayer—including a full recitation of Psalm 91. Hearing these words chanted in a living community brings the ancient Syriac fragment to life in ways that no algorithm can replicate!
Acknowledgments
I would like to express my sincere gratitude to the warm hospitality of the Turfan conference organizers: The Association of Dunhuang and Turfan Studies of China, Sun Yat-sen University, the Cultural Relics Bureau of Xinjiang Uygur Autonomous Region, the Turpan Municipal Committee, and the Mayor of Turpan City: Thank you, 谢谢 Xièxiè!
