-
Notifications
You must be signed in to change notification settings - Fork 0
Research Translation History Gemini
The Computational Renaissance of Cuneiform Studies: An Exhaustive Analysis of Digital Transformation, Artificial Intelligence Integration, and Collaborative Infrastructure in Assyriology
The study of the ancient Near East stands at a critical juncture, transitioning from a discipline defined by the painstaking manual decipherment of individual clay tablets to one powered by large-scale digital repositories and sophisticated machine learning architectures. For nearly two centuries, the field of Assyriology was characterized by a fundamental bottleneck: the volume of recovered archaeological material—estimated at more than 500,000 cuneiform artifacts—vastly exceeded the human capacity for translation and publication.1 Cuneiform, a wedge-based writing system used for over three millennia to encode languages as diverse as Sumerian, Akkadian, Elamite, and Hittite, represents the world’s oldest and most extensive record of human history.1 However, the physical fragmentation of these artifacts, dispersed across approximately 1,200 public and private collections globally, has historically hindered the synthesis of this data.5
The emergence of the Cuneiform Digital Library Initiative (CDLI) in the late 1990s and the recent launch of the Institute for the Study of Ancient Cultures (ISAC) Data Research Center in 2025 represent institutional shifts toward a "computationally meaningful" approach to antiquity.7 By integrating high-resolution imaging, standardized encoding protocols such as the ASCII Transliteration Format (ATF), and automated reconstruction tools like the Electronic Babylonian Library (eBL), the domain is moving beyond simple digitization toward a holistic digital ecosystem.8 This transformation is not merely technical but philosophical, redefining how historical knowledge is curated, shared, and repatriated through international collaborations and open-access frameworks.12
The evolution of digital cuneiform studies is rooted in the long-standing institutional efforts to catalog the vast textual remains of Mesopotamia. The University of Chicago’s Institute for the Study of Ancient Cultures, founded in 1919 as the Oriental Institute, provides a quintessential case study in this transition.8 The founder, James Henry Breasted, envisioned a systematic approach to the ancient world that mirrored the scientific rigor of the natural sciences.16
In 1933, Breasted proposed the Archaeological Corpus Project, a card-based catalog that served as the 20th-century precursor to the modern database.14 This effort reached its zenith with the Chicago Assyrian Dictionary (CAD), a monumental project initiated in 1921 and completed in 2011.15 Modeled after the Oxford English Dictionary, the CAD required nine decades to compile, providing more than just lexical equivalents by offering an exhaustive cultural and historical context for every Akkadian word.15
The limitations of this analog model were evident: the CAD was a finished, static product, whereas the archaeological record is dynamic, with new excavations constantly expanding the corpus.2 Furthermore, the physical separation of the dictionary's millions of reference cards from the actual museum artifacts created a disconnect between textual and material research.14
The digital turn began in earnest with the founding of the Cuneiform Digital Library Initiative in 1998.7 The project was led by Robert Keith Englund (UCLA) and Jürgen Renn (Max Planck Institute for the History of Science) and sought to put the estimated 500,000 recovered cuneiform tablets online.7 The initial focus was on the administrative archives of the 4th and 3rd millennia BC, the earliest and often most poorly understood witnesses to human writing.1
Funding for this phase was secured through the National Science Foundation (NSF) and the National Endowment for the Humanities (NEH), allowing the CDLI to digitize major collections at the British Museum, the Vorderasiatisches Museum, and the University of Pennsylvania.7 This era established the primary digital identifier for artifacts, the "P-number," and standardized methods for electronic capture and data archiving.5
| Milestone Year | Event / Initiative | Institutional Significance |
|---|---|---|
| 1919 | Founding of the Oriental Institute (now ISAC) | Establishment of US leadership in Near Eastern studies 15 |
| 1921 | Launch of the Chicago Assyrian Dictionary | Beginning of a 90-year effort to map the Akkadian language 16 |
| 1933 | Archaeological Corpus Project | Early conceptualization of an integrated archaeological database 14 |
| 1998 | Founding of the CDLI | Shift toward international, internet-based dissemination of cuneiform 7 |
| 2000 | NSF/NEH Digital Libraries Grant | Secured federal funding for large-scale museum digitization 7 |
| 2011 | Completion of the CAD | Transition from the analog dictionary era to digital scholarship 16 |
| 2018 | Launch of eBL project (LMU Munich) | Application of AI to text reconstruction (Fragmentarium) 11 |
| 2025 | Launch of ISAC Data Research Center | Integration of AI, data science, and humanities at UChicago 8 |
The contemporary landscape is dominated by three major institutional frameworks: the CDLI, the ISAC Data Research Center, and the Electronic Babylonian Library.
As of 2026, the CDLI remains the world’s most comprehensive digital index of cuneiform. Managed by an international directorship across the University of York, Oxford, CNRS Nanterre, and the Max Planck Institute, the CDLI has cataloged more than 400,000 artifacts.5 Its data model is built around the "artifact identifier" (P-number), which tracks the physical object, and the "composite number" (Q-number), which tracks unique textual compositions.5 This distinction is critical for scholars tracking the transmission of literary works across different archaeological sites.5
The CDLI has expanded its mission through the CDLI-ACT (Access to Cuneiform Texts) project, an Arabic-language interface launched in 2025.12 Directed by scholars from Oxford and Al-Qadisiyah University, CDLI-ACT represents a significant effort toward digital repatriation, enabling researchers in Iraq to engage with artifacts that are physically housed in European or North American museums.12
The University of Chicago's ISAC has evolved its historical "Integrated Database Project" into the Data Research Center (DRC).8 The DRC integrates ISAC’s extensive research archives, which contain over one million records including 100,000 photographic negatives documenting excavations since 1892.8 The center’s core mission is to make these records "computationally meaningful" by applying AI to the century-long Chicago Assyrian Dictionary files and millions of physical notecards.8
The DRC serves as a hub linking several specialized projects:
- The Chicago Assyrian Dictionary Digital Transformation: Converting analog lexical data into a searchable, NLP-ready database.8
- The Aqaba Glass Database and Ancient Egyptian Demonology Project: Expanding the digital methodology to non-cuneiform materials.8
- OCHRE (Online Cultural and Historical Research Environment): A data platform supporting projects like DeepScribe.22
Based at Ludwig Maximilian University of Munich (LMU), the eBL project has pioneered the use of AI for tablet reconstruction.11 Its "Fragmentarium" tool addresses the fact that many thousands of tablets remain in fragments, with pieces of the same original document often stored in different museums.23 Since 2018, eBL has processed over 22,000 fragments and discovered approximately 1,200 "joins".23 In November 2022, the software successfully identified a fragment belonging to a late version of the Gilgamesh epic dating to 130 BC, demonstrating the power of automated matching.23
The digital translation of cuneiform requires a robust technical substrate that accounts for the script’s unique wedge-shaped morphology and three-dimensional nature.3
Cuneiform was officially added to the Unicode Standard in 2006 (U+12000 block), providing a stable foundation for machine-readable text.26 This standard establishes the identity of signs through representative glyphs and names, covering Sumero-Akkadian signs, numbers, and early dynastic variants.27 However, Unicode is limited for philological research because it represents idealized forms, whereas actual cuneiform characters vary significantly by period, region, and individual scribe.3
| Unicode Range | Content Category | Functional Significance |
|---|---|---|
| U+12000–U+123FF | Sumero-Akkadian Cuneiform | Core sign list for standard Sumerian and Akkadian 27 |
| U+12400–U+1247F | Cuneiform Numbers and Punctuation | Specialized characters for administrative and accounting texts 27 |
| U+12480–U+1254F | Early Dynastic Cuneiform | Archaic signs for the earliest witnesses of writing 27 |
Because cuneiform is polyphonic (one sign can have multiple sounds) and homophonic (one sound can be written with multiple signs), scholars use the ATF standard to create machine-interpretable versions of the text.2 ATF uses ASCII characters with specialized subscripts and diacritics to distinguish values.
An ATF file is structured using several functional markers:
- &-lines: Identification of the artifact (e.g., &P000001 = ATU 3, pl. 011, W 6435,a).10
- @-lines: Object surfacing and structural markers (e.g., @obverse, @column 1, @edge).10
- $-lines: Philological asides describing the physical state of the tablet (e.g., $ broken, $ some lines missing).10
- Text lines: The actual content, which must follow strict rules for sign naming and spacing to allow for automated lemmatization.10
Traditional 2D photography often fails to capture the depth of the wedge, which is essential for distinguishing similar signs.25 Reflectance Transformation Imaging (RTI) has become the archival standard, allowing researchers to virtually move the light source across the surface of a digitized tablet.32 Furthermore, 3D modeling allows for the digital reconstruction of curved surfaces, which is particularly useful for cylinder seals and multi-faced prisms.25
The integration of AI into cuneiform studies focuses on three primary areas: computer vision (OCR), natural language processing (NLP), and neural machine translation (NMT).
The DeepScribe project, a collaboration between ISAC and the UChicago Department of Computer Science, aims to automate the localization and identification of signs on tablets from the Persepolis Fortification Archive.22 Trained on over 6,000 annotated images, the model achieves significant accuracy in identifying the Elamite language.25
| Model Component | Architecture / Method | Reported Performance |
|---|---|---|
| Sign Localization | RetinaNet | 0.78 mAP 25 |
| Sign Classification | ResNet | 0.89 Top-5 Accuracy 25 |
| End-to-End Pipeline | CNN + Morphological Clustering | 0.80 Top-5 Accuracy 31 |
DeepScribe’s innovation lies in its "hotspot" system, where over 100,000 individually identified signs were annotated by students to build a robust training set.31 This allows the computer to provide researchers with ranked probabilities for a sign's identity, accelerating the work of experts by filtering out repetitive administrative sequences.31
A critical obstacle in automated cuneiform reading is the high variability of sign forms. Researchers at Cornell and Tel Aviv University developed "ProtoSnap," which uses generative AI diffusion models to "snap" a prototype sign onto the varying strokes found on a physical tablet.3 By calculating the similarity between pixels and idealized character prototypes, ProtoSnap improves the accuracy of downstream OCR models, allowing for large-scale comparisons across different cities and writers.3
Translating cuneiform into English is a "low-resource" problem in AI terms, as there is a limited corpus of paired data for training.2 Projects like "CuneiTranslate" have experimented with transformer architectures and models like Meta's NLLB (No Language Left Behind) and T5.2
While some automated corpora (like the AICC with 130,000 AI-translated texts) exist, scholars caution that Large Language Models (LLMs) often generate "nonsense" or "hallucinations" when applied to cuneiform without rigorous constraints.35 LLMs struggle with the discontinuous morphology of Akkadian and the high degree of ambiguity in Sumerian.35 The EvaCun 2025 Shared Task, part of the Second Workshop on Ancient Language Processing, highlighted that while models can achieve 90% accuracy for in-vocabulary lemmatization, their performance drops to roughly 9% for out-of-vocabulary (OOV) terms.37
The digital transformation of Assyriology is inherently collaborative, involving academic-industry partnerships and a reliance on open-source frameworks.
The CDLI-ACT project represents a flagship collaboration between Oxford University and Al-Qadisiyah University in Iraq, funded by the British Institute for the Study of Iraq and the Mellon Foundation.12 This partnership is essential for sustainable digital heritage preservation, as it includes workshops and seminars in Iraq to train a new generation of digital humanists.12
The eBL project likewise collaborates with the Iraq Museum and the British Museum, ensuring that fragment data is digitized and shared across borders.23 Within the United States, the ISAC Data Research Center works closely with UChicago IT Services and the Research Computing Center to maintain the infrastructure for its million-record database.8
As digital cuneiform projects are often non-profit and international, they face complex financial challenges. The CDLI has adopted the "Open Collective" platform for fiscal sponsorship.5 This allows the project to raise funds transparently from individual donors and institutions without the overhead of maintaining a dedicated 501(c)(3) entity for every initiative.5 Open Collective acts as a "fiscal host," managing the legal and financial paperwork while the researchers focus on content.40
Similarly, the Zooniverse platform, which hosts crowdsourced tablet annotation projects, is supported by Chicago's Adler Planetarium, a 501(c)(3) organization.43 This model allows scientific and cultural projects to operate under a common legal umbrella, facilitating public participation and donor engagement.43
The current state of the domain offers multiple pathways for individual contributors—from domain experts and computer programmers to citizen scientists.
The CDLI and eBL provide portals for scholars to contribute directly to the data collections. Registration with the CDLI allows users to edit metadata, upload transliterations, and contribute images.5 The process is governed by strict documentation guidelines to ensure archival standards are met:
- Scanning Requirements: Tablets should be scanned at 600 dpi in a specific order (obverse, reverse, left/right edges, top/bottom edges) to create the archival "fat cross" representation.44
- Metadata Submission: Researchers can use "bulk upload" forms to submit changes to artifact records or bibliographical entries.32
- Code Contributions: Developers can contribute to the CDLI framework on GitLab (cdli/framework) or standalone projects on GitHub (cdli-gh) after signing a Contributor License Agreement.32
For non-experts, the Zooniverse platform provides an entry point into "people-powered research".43 Volunteers can participate in projects like "Deciphering Secrets," where they help transcribe handwritten historical documents or annotate subcellular structures in 3D biological imaging.43 In the context of cuneiform, this "segmentation" task—coloring in pixels to identify specific signs—is vital for training AI models like DeepScribe.45
The Zooniverse "First Look" initiative also allows the public to join in discoveries from the Legacy Survey of Space and Time (LSST), demonstrating a model of participation that could be applied to large-scale cuneiform fragment matching in the future.46
For those with technical skills, projects like the eBL API require specific local development environments. This includes:
- Docker Desktop: For containerized application management.48
- VS Code with Dev Containers: To maintain a consistent development setup.48
- MongoDB and Auth0: For database and authentication services.48
- Black Codestyle and PEP8: Standards for Python development to ensure code interoperability.48
The domain is entering a phase of rapid acceleration, characterized by the integration of "big data" and artificial intelligence into the core of the discipline.
Several major milestones are slated for the 2025–2026 period:
- CDLI Arabic Interface (CDLI-ACT): A test version is expected in late 2025, with a production version launching in the first half of 2026 at the Rencontre Assyriologique Internationale in Baghdad.12
- Thesaurus Linguarum Hethaeorum (TLHdig 1.0): Expected in late 2025, this tool will provide complete coverage of all published Hittite texts, comprising over 400,000 transliterated lines.49
- ISAC Data Research Center Expansion: The DRC is continuing the digital transformation of the Chicago Assyrian Dictionary, aiming to make it computationally searchable across different historical periods and geographic regions.8
As LLMs like GPT-4 and its successors continue to evolve, the field is moving away from simple zero-shot translation toward more robust "Retrieval-Augmented Generation" (RAG) and neuro-symbolic AI.50 By linking LLMs to structured knowledge bases like the FactGrid Wikibase (which maps Sumerian/Akkadian lexemes to all their English senses), researchers can constrain AI outputs to produce more accurate and scholarly translations.2
The most significant future trend is the integration of heterogeneous data types. The ISAC DRC is positioned to lead this effort by connecting archaeological data (geospatial modeling) with linguistic data (NLP) and environmental data (climate patterns).8 This "multi-modal" approach allows for insights at scales previously impossible, such as tracking the economic impact of climate shifts in the Old Babylonian period by analyzing thousands of administrative tablets simultaneously.8
The digital transformation of cuneiform translation and digitization has moved the field from a 19th-century philological model to a 21st-century data science paradigm. Initiatives like the CDLI and ISAC's Data Research Center have not only preserved the fragile remains of the ancient Near East but have unharnessed their content for a global audience.1 The transition from manual transcription to AI-assisted "Fragmentarium" matching represents a fundamental shift in how humanity engages with its earliest written history.6
The future success of these projects depends on a delicate balance: the technical rigor of computer science, the linguistic expertise of Assyriology, and a commitment to open-access, ethical data practices.8 By fostering partnerships across borders—exemplified by the CDLI-ACT project—and inviting participation from a global community of scholars and citizen scientists, the field is ensuring that the "forgotten" myths of the storm god Iškur or the daily business records of Old Assyrian merchants are no longer lost to time, but are readable, searchable, and meaningful for future generations.12
As AI models continue to "snap" character forms and "join" digital fragments, the digital resurrection of Mesopotamia is increasingly becoming a reality, transforming ancient clay into a vibrant, modern database of human experience.3
- Cuneiform Digital Library Initiative (CDLI) - Faculty of Asian and Middle Eastern Studies, accessed February 3, 2026, https://www.ames.ox.ac.uk/cuneiform-digital-library-initiative-cdli
- Cuneiform Translation Project, accessed February 3, 2026, https://anniepang.github.io/CuneiformTranslationWebsite/
- AI Could Translate 5000-Year-Old-Language, Saving Time and Historical Insights, accessed February 3, 2026, https://www.discovermagazine.com/ai-could-translate-5-000-year-old-language-saving-time-and-historical-47208
- EvaCun 2025 Shared Task : Lemmatization and Token Prediction in Akkadian and Sumerian using LLMs Gordin - Helda - University of Helsinki, accessed February 3, 2026, https://helda.helsinki.fi/bitstreams/bd231abd-506a-4e66-8803-1620a4d06297/download
- The Cuneiform Digital Library Initiative: A Primer, accessed February 3, 2026, https://iaassyriology.com/the-cuneiform-digital-library-initiative-a-primer/
- Lost for 3,000 Years, an Ancient Babylonian Hymn Has Been Restored Using Artificial Intelligence - The Debrief, accessed February 3, 2026, https://thedebrief.org/lost-for-3000-years-an-ancient-babylonian-hymn-has-been-restored-using-artificial-intelligence/
- Cuneiform Digital Library Initiative - Wikipedia, accessed February 3, 2026, https://en.wikipedia.org/wiki/Cuneiform_Digital_Library_Initiative
- ISAC launches Data Research Center to advance digital discovery in the humanities, accessed February 3, 2026, https://news.uchicago.edu/story/isac-launches-data-research-center-advance-digital-discovery-humanities
- ISAC launches Data Research Center | Institute for the Study of ..., accessed February 3, 2026, https://isac.uchicago.edu/article/isac-launches-data-research-center
- The Open Richly Annotated Cuneiform Corpus - ATF Structure Tutorial - Oracc, accessed February 3, 2026, https://oracc.museum.upenn.edu/doc/help/editinginatf/primer/structuretutorial/
- The “electronic Babylonian Library” (eBL) Platform, accessed February 3, 2026, https://www.ebl.lmu.de/
- Access to Cuneiform Texts (CDLI-ACT) - the Cuneiform Digital Library Initiative in Arabic, accessed February 3, 2026, https://cdli.earth/postings/225
- Cuneiform Digital Library Initiative: Home, accessed February 3, 2026, https://cdli.earth/
- ISAC's New Data Research Center Brings the Ancient World Into the Digital Age, accessed February 3, 2026, https://chicagomaroon.com/50070/news/isacs-new-data-research-center-brings-the-ancient-world-into-the-digital-age/
- Chicago Assyrian Dictionary - Wikipedia, accessed February 3, 2026, https://en.wikipedia.org/wiki/Chicago_Assyrian_Dictionary
- The Chicago Assyrian Dictionary (CAD) - atour.com, accessed February 3, 2026, https://www.atour.com/library/cad/
- INTEGRATED DATABASE PROJECT - Institute for the Study of Ancient Cultures, accessed February 3, 2026, https://isac.uchicago.edu/sites/default/files/uploads/shared/docs/Publications/Annual-Reports/2019-2020/AR2019-20_IDB.pdf
- Cuneiform Digital Library Initiative (CDLI) - OMNIKA, accessed February 3, 2026, https://omnika.org/library/cuneiform-digital-library-initiative-cdli
- Cuneiform Digital Library Initiative Project (CDLI) - MPIWG, accessed February 3, 2026, https://www.mpiwg-berlin.mpg.de/project_cdli
- About CDLI - Cuneiform Digital Library Initiative, accessed February 3, 2026, https://cdli.earth/about
- Collections - Institute for the Study of Ancient Cultures - The University of Chicago, accessed February 3, 2026, https://isac.uchicago.edu/collections/collections
- DeepScribe | The Online Cultural and Historical Research ..., accessed February 3, 2026, https://voices.uchicago.edu/ochre/project/deepscribe/
- Electronic Babylonian Literature: Playing with the source of world literature - LMU München, accessed February 3, 2026, https://www.lmu.de/en/newsroom/news-overview/news/electronic-babylonian-literature-playing-with-the-source-of-world-literature-4d9706d5.html
- About - The “electronic Babylonian Library” (eBL) Platform, accessed February 3, 2026, https://www.ebl.uni-muenchen.de/about
- DeepScribe: Localization and Classification of Elamite Cuneiform Signs Via Deep Learning, accessed February 3, 2026, https://www.researchgate.net/publication/389664311_DeepScribe_Localization_and_Classification_of_Elamite_Cuneiform_Signs_Via_Deep_Learning
- 𒀀 - U+12000 - decodeunicode.org, accessed February 3, 2026, https://decodeunicode.org/en/u+12000
- Cuneiform (Unicode block) - Wikipedia, accessed February 3, 2026, https://en.wikipedia.org/wiki/Cuneiform_(Unicode_block)
- UTR #56: Unicode Cuneiform Sign Lists, accessed February 3, 2026, https://www.unicode.org/reports/tr56/
- AI models make precise copies of cuneiform characters | Cornell Chronicle, accessed February 3, 2026, https://news.cornell.edu/stories/2025/03/ai-models-make-precise-copies-cuneiform-characters
- The Open Richly Annotated Cuneiform Corpus - ATF Inline Tutorial - Oracc, accessed February 3, 2026, https://oracc.museum.upenn.edu/doc/help/editinginatf/primer/inlinetutorial/index.html
- DeepScribe AI Can Help Translate Ancient Tablets - Unite.AI, accessed February 3, 2026, https://www.unite.ai/deepscribe-ai-can-help-translate-ancient-tablets/
- Contribute - Cuneiform Digital Library Initiative, accessed February 3, 2026, https://cdli.earth/contribute
- How AI could help translate the written language of ancient civilizations - UChicago News, accessed February 3, 2026, https://news.uchicago.edu/story/how-ai-could-help-translate-written-language-ancient-civilizations
- CuneiTranslate: Unlocking Ancient Mesopotamian Knowledge, accessed February 3, 2026, https://www.ischool.berkeley.edu/projects/2024/cuneitranslate-unlocking-ancient-mesopotamian-knowledge
- World's Largest Translated Cuneiform Corpus using AI - Reddit, accessed February 3, 2026, https://www.reddit.com/r/Cuneiform/comments/1g4fdig/worlds_largest_translated_cuneiform_corpus_using/
- Linguistic annotation of cuneiform texts using treebanks and deep learning - ResearchGate, accessed February 3, 2026, https://www.researchgate.net/publication/377954291_Linguistic_annotation_of_cuneiform_texts_using_treebanks_and_deep_learning
- EvaCun 2025 Shared Task: Lemmatization and Token Prediction in Akkadian and Sumerian using LLMs - ACL Anthology, accessed February 3, 2026, https://aclanthology.org/2025.alp-1.33/
- EvaCun 2025 Shared Task: Lemmatization and Token Prediction in Akkadian and Sumerian using LLMs - University of Helsinki Research Portal, accessed February 3, 2026, https://researchportal.helsinki.fi/en/publications/evacun-2025-shared-task-lemmatization-and-token-prediction-in-akk/
- A Low-Shot Prompting Approach to Lemmatization in the EvaCun 2025 Shared Task, accessed February 3, 2026, https://aclanthology.org/2025.alp-1.31/
- Open Collective Blog, accessed February 3, 2026, https://blog.opencollective.com/
- News - Open Collective, accessed February 3, 2026, https://blog.opencollective.com/tag/news/
- Fiscal Sponsors. We need you! - Open Collective, accessed February 3, 2026, https://blog.opencollective.com/fiscal-sponsors-we-need-you/
- April | 2025 | Zooniverse, accessed February 3, 2026, https://blog.zooniverse.org/2025/04/
- Contribute - CDLI Wiki, accessed February 3, 2026, https://cdli.ox.ac.uk/wiki/doku.php?id=submission_guidelines2
- October | 2025 | Zooniverse, accessed February 3, 2026, https://blog.zooniverse.org/2025/10/
- June | 2025 | Zooniverse, accessed February 3, 2026, https://blog.zooniverse.org/2025/06/
- DISCOVER ZOONIVERSE PROJECTS IN A WHOLE NEW WAY, accessed February 3, 2026, https://blog.zooniverse.org/2026/01/27/discover-zooniverse-projects-in-a-whole-new-way/
- ElectronicBabylonianLiterature/ebl-api: Electronic Babylonian Literature API - GitHub, accessed February 3, 2026, https://github.com/ElectronicBabylonianLiterature/ebl-api
- Cuneiforms: New digital tool for translating ancient texts | ScienceDaily, accessed February 3, 2026, https://www.sciencedaily.com/releases/2025/03/250326123733.htm
- Proceedings of the First Workshop on Comparative Performance Evaluation: From Rules to Language Models - ACL Anthology, accessed February 3, 2026, https://aclanthology.org/2025.r2lm-1.pdf
- UChicago Sumerologist translates forgotten 4400-year-old myth, accessed February 3, 2026, https://news.uchicago.edu/story/uchicago-sumerologist-translates-forgotten-4400-year-old-myth
- Neural Machine Translation of Old Assyrian Cuneiform Business Records into English, accessed February 3, 2026, https://www.researchgate.net/publication/400071542_Neural_Machine_Translation_of_Old_Assyrian_Cuneiform_Business_Records_into_English
Source: github.com/wittkensis/glintstone · Issues · Edit this wiki
Start here
Getting Started
Overview
Data Model
- Data Sources
- Data Quality
- Data Issues
- Import Pipeline Guide
- ML Integration
- Citation Pipeline Summary
Reference — Data Model
Reference — API
Reference — MCP
Opportunities
Personas
Project
Research