Sphragis

Sphragis (σφραγίς, "sigil") is a benchmark for Ancient Greek (grc) prose and verse authorship attribution (AA). Its input is the complete curated union of the human-annotated CoNLL-U trees in the supported treebank projects, including lemma, part of speech, morphology, and dependency relations. It defines 1-, 10-, and 100-sentence attribution tasks. The complementary scanned-line benchmark is published separately as Urdatorn/sphragis-metre.

For the archaic Greek poets, "sphragis" (or "sphregis" in some dialects) was a term of art roughly equivalent to the modern "authorial watermark", referring to text composition that intentionally aims at making it harder for adversaries to falsely pass off the work as their own. The phenomenon of "sphragizing" is one of the earliest chapters in the history of AA. (See Theognidea, 19-30)

Sphragis does not require a sentence to occur in any second annotation stream. Prose and verse enter on the same basis: a retained sentence from any supported treebank is eligible. Exact duplicates are consolidated with all treebank provenance preserved.

Tasks

There are three chunk sizes. Prose and verse sentences share one closed-set authorship problem; the genre column preserves their source representation.

Configurations

Configuration	Authors	Row unit	Train rows	Validation rows	Test rows
sentence_1	17	sentence	43,300	4,800	4,800
sentence_10	17	10-sentence chunk	4,330	480	480
sentence_100	17	100-sentence chunk	433	48	48

The suffix gives the chunk size in every split. _100 defines one shared source corpus for all three tasks: an author is retained only when that author independently has at least 100 source rows in both validation and test, and every split is reduced to a multiple of 100 source rows per author. _1 publishes the shared rows atomically, while _10 and _100 regroup those exact same train, validation, and test rows into fixed benchmark chunks. Thus author inventory and total source text are identical across task sizes.

Source-row selection uses seed 776. For each retained author and split, the unavoidable n mod 100 source rows are selected for exclusion once by a stable hash. The shared retained rows are put into their best available textual order within each work for each chunking task. Complete single-work chunks are emitted first; only residual work tails are combined, and such rows are marked by chunk_is_mixed_work. Chunk IDs, boundaries and exclusions are therefore reproducible. The complete manifest is in metadata/dataset_variants.json.

Every row retains its treebank genre and provenance.

from datasets import load_dataset

sentences = load_dataset("Urdatorn/sphragis", "sentence_10")

Splits

Splits use random seed 776 and begin with independent stratification within each author/work group: approximately 80% train, 10% validation and 10% test by row. Groups with 3–9 rows receive one validation and one test row; groups with fewer than three stay in training. The 100-row eligibility and remainder bottleneck is then applied to all three task sizes. The train, validation, and test source sets are all exact multiples of 100 per author. This makes the published source-row ratio slightly different from the initial 80/10/10 split. A stricter work-held-out evaluation can be added as a later track.

Principal columns

All configurations contain:

• id, author, work, work_id, genre, text;
• conllu, cts_urn, passage, treebank_source;
• source_records: JSON containing every contributing upstream record, revision, URL, license, annotation provenance and syntax scheme;
• licenses, dedup_key, and split.

text is always a JSON list of strings. Atomic _1 rows contain one string; every split of _10 and _100 contains exactly 10 or 100 separate sentence strings. No whitespace separator is used to encode unit boundaries.

conllu is likewise always a JSON list of strings with the same number of units as text.

The _10 and _100 configurations additionally contain chunk_size, chunk_target_size, ordered constituent_ids, JSON constituent_provenance, chunk_work_ids, chunk_works, and chunk_is_mixed_work. Every row has chunk_size equal to the configuration suffix. Aggregated text and conllu follow constituent order, while constituent provenance retains the original passage-level locations.

All sentence rows share one schema. There are no metrical, scanned-line, or external-alignment columns. genre has exactly two values: prose and verse.

Dependency representation

Existing UD data is retained verbatim. Native AGDT/Arethusa trees are serialized as CoNLL-U while preserving heads. Their dependency relations are conservatively mapped to the Universal Dependencies v2 inventory; opaque source-specific categories fall back to dep. The lossless original NativeRel and NativeHead values remain in MISC for audit and future mapping improvements. Repairs required after removing artificial ellipsis nodes—or to break an upstream beta self-loop/cycle—are explicitly marked HeadRepair=Yes.

XPOS is normalized across every source to the nine-position Ancient Greek AGDT/Perseus tag convention. Existing valid positional tags are retained; incompatible source-specific tags are reconstructed from UPOS and FEATS, with unavailable distinctions represented by -. UPOS and FEATS remain the canonical UD fields.

Every distinct published conllu document is loaded during testing with the official CoNLL 2018 shared-task load_conllu parser. This checks sentence termination, ten-column structure, token IDs, heads, roots and dependency cycles across all configurations.

To prevent authorship-label leakage, model-facing conllu values retain only the non-identifying # text = ... comment. Comments carrying sent_id, source, cts, passage, or newdoc id are removed. The same information remains available for audit in the dataset's dedicated provenance columns. Token-level MISC identifiers (Cite, Ref, and LId) are also removed; only NativeRel, NativeHead, HeadRepair, and SpaceAfter are retained there.

Deduplication

Raw treebank input comprised 110,849 prose and verse sentences. Authorship curation retained 92,141 source sentences before deduplication. Exact normalization-based deduplication then removed 23,282 repeated rows, and 1,291 rows in 253 exact-text groups were omitted because their retained author labels conflict, leaving 67,568 eligible sentences before the shared 100-row task bottleneck. Normalization lowercases, maps final sigma, removes diacritics and retains letters only. The chosen parse follows a deterministic source priority, while all exact-source provenance is merged into source_records.

This is aggressive exact deduplication, not passage-level fuzzy deduplication. Segmentation or edition differences can therefore leave near duplicates for future review.

Sources and licensing

The corpus draws on AGDT 2.1, UD Perseus, UD PROIEL, UD PTNK, Gorman's Greek Dependency Trees, Pedalion, and Harrington Trees.

This is a mixed-license dataset, including NonCommercial and ShareAlike data. Consult LICENSES.md and the licenses recorded on every row. Source commits for this release are frozen in metadata/source_revisions.json.

Conservative authorship policy

The benchmark favors reliable training labels over coverage. It excludes anonymous and unknown authors, Pseudo-* and traditional attributions, fragment-labelled works, corporate or mediated authorship, and works or passages whose attribution is substantially disputed. This includes disputed material transmitted inside the Demosthenic, Antiphontic, Lysian, Platonic, Aeschylean, Isocratean and Xenophontic corpora, as well as the pseudo-Hesiodic Shield of Heracles. The core Theogony and Works and Days are retained under Hesiod. Romans is retained under Paul, while disputed and anonymous New Testament works are excluded.

The Iliad and Odyssey are retained as distinct conventional corpora under the author labels Homeric-Iliad and Homeric-Odyssey; the labels do not claim a shared biographical author. Every work labelled Fragments is excluded. Machine-readable reasons and pre-deduplication counts are recorded in metadata/build_report.json.