PoetryMTEB/LatinEpicIntertextualityRetrieval

Latin Epic Intertextuality Retrieval

BEIR-style Retrieval benchmark for Latin epic intertextuality under PoetryMTEB.

Given a passage from Valerius Flaccus, Argonautica Book 1, retrieve the corresponding verse line(s) in Vergil (Aeneid), Lucan (Bellum Civile), Ovid (Metamorphoses), or Statius (Thebaid) that traditional scholarship identifies as parallels.

Gold parallels: Burns et al., NAACL-HLT 2021 (paper; repo), 945 curated pairs.

Dataset Card

Item	Description
Dataset version	1.0.0
Task	Retrieval (document ranking)
Language	Latin (la)
Query	VF Argonautica 1.x line span (queries.text)
Corpus	Line-level epic verses (Vergil / Lucan / Ovid / Statius)
Relevance	Scholar-curated parallels (score=1); multi-line targets → all lines marked relevant
Splits	test only (MTEB Retrieval)
Size	corpus=39675; queries=482; qrels=978
Avg. relevant docs / query	2.03 (min=1, max=10)
License	CC BY 4.0 (packaging; classical texts are public domain)
Evaluation metrics	nDCG@10, MAP, Recall@k

Corpus author distribution (documents)

author	work	#lines
Ovid	Metamorphoses	11995
Vergil	Aeneid	9896
Statius	Thebaid	9731
Lucan	Bellum Civile	8053

Configs

corpus

Field	Type	Description
id	string	{author_key}-{book}-{line} (e.g. vergil-1-1)
text	string	Retrieval document (Latin verse line)
author	string	Vergil / Lucan / Ovid / Statius
work	string	Work title
book	int64	Book number
line	int64	Line number
tess_index	string	Original Tesserae-style index

queries

Field	Type	Description
id	string	vf-1-{start}-{end}
text	string	Retrieval query: concatenated VF Book 1 lines
vf_book	int64	Always 1 in this release
line_start / line_end	int64	Inclusive VF line span
query_phrases	string	JSON list of lemma/phrase queries from the gold file
n_parallels	int64	Number of gold parallels for this span

qrels

Field	Type	Description
query-id	string	Matches queries.id
corpus-id	string	Matches corpus.id
score	float64	1.0 for gold relevant lines

Construction method

1. Load Tesserae-format .tess line files for four target epics → corpus.
2. Load VF Argonautica Book 1 lines; group gold rows in vf_intertext_dataset_1_0.csv by (VF Line Start, VF Line End).
3. Query text = concatenated VF lines in the span.
4. For each parallel, mark every target line in [Intertext Line Start, End] as relevant.
5. Deduplicate qrels per (query-id, corpus-id).

How to load

from datasets import load_dataset

corpus = load_dataset("PoetryMTEB/LatinEpicIntertextualityRetrieval", "corpus")["test"]
queries = load_dataset("PoetryMTEB/LatinEpicIntertextualityRetrieval", "queries")["test"]
qrels = load_dataset("PoetryMTEB/LatinEpicIntertextualityRetrieval", "qrels")["test"]

print(queries[0]["id"], queries[0]["text"][:120])
print(corpus[0]["author"], corpus[0]["text"])

Citation

@inproceedings{burns-etal-2021-profiling,
  title     = {Profiling of Intertextuality in {L}atin Literature Using Word Embeddings},
  author    = {Burns, Patrick J. and Brofos, James A. and Li, Kyle and Chaudhuri, Pramit and Dexter, Joseph P.},
  booktitle = {Proceedings of NAACL-HLT 2021},
  year      = {2021},
  pages     = {4900--4907},
  url       = {https://aclanthology.org/2021.naacl-main.389/}
}

Upstream: https://github.com/QuantitativeCriticismLab/NAACL-HLT-2021-Latin-Intertextuality
Hub packaging: PoetryMTEB/LatinEpicIntertextualityRetrieval (v1.0.0)