parisaslanidis/eleni

Ελένη: Digital Corpus of Greek Parliamentary Proceedings, 1862–1967

Version v1.0 · 2026-11-06 · the corpus behind eleni.ekke.gr

Ελένη is an AI-ready digital corpus of the debates of the Greek Parliament, from 1862 (the Second National Assembly) to 1967 (the Eighth Period). Every sitting and every printed row has a stable address on the site, and the same addresses are given here, so whatever is found in the dataset can be read, checked and cited on the site.

Η Ελένη είναι ένα AI-ready ψηφιακό αποθετήριο πολιτικού λόγου: οι συζητήσεις του Ελληνικού Κοινοβουλίου από το 1862 (Β΄ Εθνική Συνέλευσις) έως το 1967 (Περίοδος Η΄), με σταθερή διεύθυνση για κάθε συνεδρίαση, παρέμβαση και στίχο.

Sittings	7,294
Turns (interventions by speakers)	1,175,931
Rows (printed lines)	16,305,278
Words	122,619,166
Sessions	107 present, of 118 known (27 periods)

Source

The stenographic record of the debates, published first as the Επίσημος Εφημερίς της Εθνοσυνελεύσεως and the Επίσημος Εφημερίς της Συνελεύσεως, and from 1865 to 1967 as the Εφημερίς των Συζητήσεων της Βουλής. (Until 1974 the Πρακτικά των Συνεδριάσεων της Βουλής were a separate publication and are not included.) The volumes were digitised from scratch: images from the Library of the Hellenic Parliament (CC BY 4.0), new scans of donated and purchased copies, and volumes photographed in libraries across Greece. The text was read with AI models (Google Gemini 3/3.1 Pro), processed in Python, and corrected by hand; the turns (who speaks where) were found automatically and reviewed volume by volume.

The text is in the language and spelling of its time: mostly katharevousa, in the polytonic script, as printed.

The two configurations

from datasets import load_dataset

sittings = load_dataset("parisaslanidis/eleni", "sittings", revision="v1.0", split="train")
print(sittings[0]["text"][:500])      # the running text
print(sittings[0]["rows"][:5])        # the edition, line by line

The files are split by decade (data/sittings/1880s.parquet …), so a period can be read on its own, e.g. with pandas: pd.read_parquet("hf://datasets/parisaslanidis/eleni@v1.0/data/sittings/1880s.parquet").

sittings (default) — one row per sitting

Each sitting's text in the two forms the site gives it: the edition, line by line as printed, and the running text, ready for reading or analysis.

field
sitting	the sitting's address: 1882.02.13, or 1948.05.10~17 when two sittings share a date
date, year	the sitting's date
period, session, session_seq	its parliamentary period and session (σύνοδος); session_seq is the session's place in the contents (45, or 44.1 for one found later)
sitting_number	the sitting's number, as printed
url	the sitting on the site
presidency, source, source_volume, published	who presided; the gazette's title, its volume and year of publication
rows	the edition: every printed line, in order, as printed (hyphenation at the line's end included)
row_labels	each row's label, the number a citation uses (see Citing)
withdrawn_rows	labels of rows removed after publication
text	the running text: the same rows joined, one paragraph per speaker's turn, a blank line between turns, no numbers; tables and lists kept between [TABLE] and [/TABLE]. It is what the site's «running text» download gives, without that file's header
n_rows, n_turns, words, tables	the sitting's rows, turns (interventions by speakers), words and tables

The turns are where the running text starts a new paragraph: who speaks there is the name printed at its start (ΠΡΟΕΔΡΟΣ., Δ. ΖΩΤΟΣ.), as in the edition.

sessions — one row per session

Every session in the contents, the missing ones included (status: complete, partial or missing), with its parts (the gazette's volumes), dates, number of sittings and a bibliographic reference.

Citing

Every passage has a permalink on the site, e.g. https://eleni.ekke.gr/1865.09.30/12-20 (rows 12 to 20 of the sitting of 30 September 1865), and a bibliographic reference that names it, as the site gives it (Copy Reference):

Ελένη: Period Α, Session Α, Sitting 44 (30 September 1865), rows 12–20. Available at https://eleni.ekke.gr/1865.09.30/12-20 (accessed: …).

Rows are the printed lines. Permalinks are permanent: a later correction of a row's text changes no address, and rows recovered later are given labels (56.001, 56.002 after row 56) while rows removed are withdrawn, never renumbered. row_labels and withdrawn_rows are those labels. A passage found in the dataset is cited by its permalink and reference on the site; Ελένη as a whole, or the dataset, as below.

To cite the dataset, give its version (the text is corrected over time; each version is fixed). The same reference is given on the site's About page:

Aslanidis, P. (2026). Ελένη: Digital Corpus of Greek Parliamentary Proceedings, 1862–1967 (Version 1.0) [Data set]. National Centre for Social Research (EKKE). https://doi.org/10.57967/hf/11475

@misc{aslanidis_eleni,
  author    = {Aslanidis, Paris},
  title     = {Ελένη: Digital Corpus of Greek Parliamentary Proceedings, 1862–1967},
  year      = {2026},
  publisher = {National Centre for Social Research (EKKE)},
  version   = {v1.0},
  url       = {https://huggingface.co/datasets/parisaslanidis/eleni},
  doi       = {10.57967/hf/11475},
  note      = {https://eleni.ekke.gr}
}

Versions

The corpus is corrected and extended over time. Every release is a tagged version of this repository (v1.0, v1.1 …) and stays available: load a fixed one with revision="v1.0", and give it when citing. main is always the latest. Each release lists the sittings whose text, rows or segmentation changed (changes/<version>.csv), and manifests/<version>.json holds a fingerprint of every sitting, so two versions can be compared sitting by sitting.

• a minor version (v1.1): corrections and additions;
• a major version (v2.0): a change in the fields or files themselves.

Limitations

• The text was read from 19th- and 20th-century print; errors of reading remain, and are corrected as they are found.
• Speakers are given as printed: one person can appear under several spellings or titles, and a title (ΠΡΟΕΔΡΟΣ, ΥΠΟΥΡΓΟΣ …) under several people.
• Some sittings are missing from the surviving volumes (see sessions).

Licence and credits

CC BY 4.0. Ελένη was created by Paris Aslanidis at the National Centre for Social Research (EKKE), within the Marie Skłodowska-Curie Actions postdoctoral project GREEKPOP (101154365), funded by the European Union's Horizon Europe programme. Contact: paslanidis at ekke.gr. ORCID: 0000-0002-9454-7001.

Changelog

v1.0 — 2026-11-06

7,294 sittings · 1,175,931 turns · 16,305,278 rows · 122,619,166 words.

First release.