Co-located with ACL 2027 in August 2027
@Kyoto, Japan

One-day workshop; exact date to be announced.

Ancient languages preserve humanity’s earliest recorded intellectual and cultural achievements, yet they remain among the most challenging targets for language technologies. Progress is constrained by structural difficulties intrinsic to philological research: scarce and non-growing corpora, damaged and fragmentary textual artefacts, limited expert annotation concentrated in a small number of institutions, absence of native speakers, and non-Latin, non-alphabetic writing systems with unstable Unicode and input conventions. Meaningful advances require methods tailored for low-resource, high-expertise multimodal scenarios, alongside evaluation metrics that distinguish genuine philological progress from fluent yet unverifiable model outputs.

ALP 2027 is the third edition of the Ancient Language Processing workshop series. It builds upon ALP 2023 (co-located with RANLP 2023, Varna, Bulgaria) and ALP 2025 (co-located with NAACL 2025, Albuquerque, USA). It complements related events including Ancient Language Translation (ALT), ML4AL, and the bi-yearly LT4HALA workshop series. ALP covers ancient languages and scripts from the emergence of writing around 3400 BCE up to 800 CE across global civilizations.

This one-day workshop will bring together computational linguists, philologists, archaeologists, digital-humanities researchers, and heritage-sector practitioners. We invite original research papers, resource descriptions, position papers, and shared-task system papers. Accepted papers will appear in the ACL Anthology proceedings.

Themes

We invite submissions organized around four primary themes, together with additional related topics:

1. Evaluation under scarcity: Benchmarks, metrics and shared-task design for small, contested test sets derived from scholarly reconstructed text editions.
2. Multimodal and visual processing of scripts: OCR / handwritten-text recognition for complex ancient scripts, sign detection and segmentation, palaeographic analysis, joint modelling of texts and their physical carrier artefacts.
3. Large and small models within philological workflows: Instruction-tuned and agent-based pipelines for lemmatisation, text restoration, dating and translation; comparisons with task-specific small-scale models; emphasis on output verifiability and provenance of machine-generated readings.
4. Data governance and reproducibility: Licensing, linked open data, minimal-computing approaches, and data bias for research working on the cultural heritage of external communities.

Additional topics of interest include but are not limited to: character-set and Unicode challenges; transliteration and input methods; tokenisation; morphological, syntactic and semantic analysis; linguistic linked open data; machine translation; corpus building and annotation; handling fragmentary and uncertain readings; NLP applications for literary and historical philology.

Languages of interest

Languages of interest span Mesopotamia, Iran, the Levant, Anatolia, Egypt, the Mediterranean, Arabia, South Asia, East Asia and Mesoamerica, including but not limited to Sumerian, Akkadian, Ancient Egyptian, Hittite, Ancient Greek, Latin, Sanskrit, Literary Chinese, Old Tibetan, Old Japanese, and Mayan.

Shared Tasks

EvaCun 2027

End-to-end OCR for cuneiform languages. Participants produce full digital transliteration from composite 2-D tablet imagery, focusing primarily on second-to-first-millennium-BCE Akkadian with optional extensions to Sumerian. Two tracks are offered: a VLM-only track and an open-architecture track. Evaluation metrics include CER, WER, exact-match and segmentation-quality scores, plus downstream MT-based secondary assessment.

EvaHan 2027

Ancient-Chinese text-difficulty grading, the first shared task for classifying classical-Chinese prose and poetry into difficulty levels. EvaHan is an established evaluation series for NLP tools for pre-modern Chinese.

Submission Types & Format Requirements

All submissions must follow the official ACL LaTeX style files. Submissions are double-blind reviewed; author information must be omitted from the submitted PDF.

• Long papers: Original mature research work — maximum 8 pages of main content; unlimited pages for references and appendices.
• Short papers: Work-in-progress, position papers, resource descriptions, pilot studies — maximum 4 pages of main content; unlimited pages for references and appendices.
• Shared-task system-description papers: Reports describing systems submitted to EvaCun 2027 or EvaHan 2027. Page limits follow short-paper guidelines.

Submissions must describe unpublished original work and must not be under concurrent submission to other venues. Accepted papers will be presented as oral presentations or posters within the one-day workshop programme, and will appear in the ACL Anthology proceedings.

Submission

The ALP 2027 submission portal will be announced here. See the submission guidelines for format and review requirements.

Important Dates (preliminary)

These are placeholder dates from the first-call draft. Final dates and exact days will be announced. All deadlines are at 11:59 PM Anywhere-on-Earth (AoE, UTC−12:00).

Diversity & Inclusion

The workshop welcomes contributions from researchers across computer-science groups, humanities departments, archaeological missions, museums and independent research teams worldwide. Our programme committee spans Europe, East Asia, North America and the Middle East, combining computational linguists and domain philologists. We encourage submissions from early-career researchers, graduate students, and scholars from communities traditionally under-represented in ACL-affiliated events. Shared-task participation is open to unaffiliated teams, and shared-task datasets are released under open licences.

Contact

For inquiries, please contact the organizing committee at alp-workshop-organizers-2027@googlegroups.com.