histde/dta-documents

Deutsches Textarchiv (DTA) Documents

This datasets hosts all documents from the Deutsches Textarchiv (DTA).

One row per work of the Deutsches Textarchiv (DTA), built from the official DTA dump dta_komplett_2026-02-10 (TCF format). The dataset contains 5,480 documents spanning 1472 to 1987 with about 205M tokens and 1.4B characters of German text, with two text views per document:

• text: the historical, layout-faithful transcription (line breaks, long s ſ, combining diacritics, original spelling);
• text_normalized: the same text in modern orthography, produced by applying the DTA CAB orthography corrections (operation="replace") to the token layer.

Rich bibliographic metadata comes directly from the embedded TEI header: authors with GND identifiers, original publication date and place, publisher, holding library and shelfmark, typeface, four genre schemes, DTA subcorpus labels and the per-work license. Licenses vary across the corpus, so please check the license_family column before reuse (see Licensing).

Dataset structure

Configs and splits

Config	Split	Rows	Files
documents (default)	train	5,480	documents/train-0000[0-5].parquet (6 shards, zstd-compressed Parquet)

There is only one split, because the dataset is a corpus and not a benchmark.

Loading

from datasets import load_dataset

ds = load_dataset("histde/dta-documents", "documents", split="train")
print(ds[0]["title"], ds[0]["year"])

# Only permissively licensed works, as a streaming iterator
permissive = load_dataset("histde/dta-documents", "documents", split="train", streaming=True)
permissive = permissive.filter(lambda d: d["license_family"] in {"cc0-1.0", "cc-by-4.0", "cc-by-3.0"})

With Polars, straight from the Parquet shards:

import polars as pl

df = pl.read_parquet("hf://datasets/histde/dta-documents/documents/train-*.parquet")
df.filter(pl.col("year") < 1700).select("id", "title", "license_family", "num_tokens")

Schema

Column	Type	Description	Example
id	string	DTA directory name, used as the primary document identifier (DTADirName). Falls back to the TCF file name	goethe_faust01_1808
dta_id	string	Numeric DTA identifier (DTAID)	200006
urn	string	Persistent URN of the work	urn:nbn:de:kobv:b4-200006-1
url	string	Landing page of the work at deutschestextarchiv.de	https://www.deutschestextarchiv.de/343016
title	string	Main title(s) from the TEI title statement. Multiple main titles are joined with \n	Iphigenie auf Tauris
subtitle	string	Subtitle(s), joined with \n. Null if absent	Ein Schauspiel
volume	string	Volume title (text of <title type="volume">). Null if absent	Dritter Theil
part	string	Part title (text of <title type="part">). Null if absent	Die sentimentalischen Dichter
authors	list<struct>	Authors as {surname, forename, gnd} structs. gnd is the GND/PND URI, may be null. Empty list for anonymous works	[{"surname": "Reideburg", "forename": "Christoph von", "gnd": "http://d-nb.info/gnd/128862580"}]
date_published	string	Raw publication date string of the original print, as given in the header	1642
year	int64	First four-digit number parsed from date_published. Null if none found	1642
place_of_publication	string	Place of publication of the original print	Breslau
publisher	string	Publisher name(s), joined with \n	Georgius Baumann
edition	string	Edition statement (text, or the n attribute if the element has no text)	2. Auflage
num_pages	int64	Page count of the original source (measure type="pages")	72
repository	string	Holding library of the digitized copy	Universitätsbibliothek Breslau
shelfmark	string	Shelfmark of the copy at the repository	Universitätsbibliothek Breslau, 4 A 277/11 / 343016
typeface	string	Typeface of the original print	Fraktur
bibl	string	Short bibliographic citation string	Reideburg, Christoph von: Kurtze Anleitung: Wie die jetzige böse Zeit/ ... Breslau, 1642.
genre_dtamain	string	Main genre, DTA scheme	Gebrauchsliteratur
genre_dtasub	string	Sub genre, DTA scheme	Leichenpredigt
genre_dwds1main	string	Main genre, DWDS scheme	Belletristik
genre_dwds1sub	string	Sub genre, DWDS scheme	Prosa
dta_corpus_labels	list<string>	DTA (sub)corpus membership labels (DTACorpus class codes)	["ready", "core"]
language	string	ISO 639-3 code of the primary language	deu
language_note	string	Human-readable language label from the header	(Früh-)Neuhochdeutsch
license	string	Per-work license URL exactly as given in the header. Varies across the corpus	http://creativecommons.org/licenses/by-sa/3.0/de/
license_family	string	Normalized license identifier derived from license (http/https, /de/, deed variants collapsed)	cc-by-sa-3.0
num_images	int64	Number of page images (DTA extent measure)	72
num_tokens	int64	Number of tokens (DTA extent measure)	11515
num_types	int64	Number of word types (DTA extent measure)	3440
num_characters	int64	Number of characters (DTA extent measure)	78795
num_sentences	int64	Sentence count from the full/ sentence layer. Null if the work has no full/ file	842
text	string	Historical transcription. Layout-faithful <text> layer from simple/ (line breaks, long s ſ, combining diacritics), or space-joined tokens if only full/ exists	Kurtze Anleitung:\nWie die jetzige boͤſe Zeit/ darinnen zwar fuͤr ſich ſelbſt/ ...
text_source	string	Origin of text: layout (from simple/) or tokens (reconstructed from the full/ token layer)	layout
text_normalized	string	Modern orthography: full/ token layer with CAB replace corrections applied, space-joined. Null if the work has no full/ file	Kurze Anleitung : Wie die jetzige böse Zeit / darinnen zwar für sich selbst / ...

Example row

{
  "id": "343016",
  "dta_id": "200006",
  "urn": "urn:nbn:de:kobv:b4-200006-1",
  "url": "https://www.deutschestextarchiv.de/343016",
  "title": "Kurtze Anleitung: Wie die jetzige böse Zeit/ darinnen zwar für sich selbst/ nichts/ alß eytel Klag/ Ach/ vnd Weh regieret",
  "authors": [{"surname": "Reideburg", "forename": "Christoph von", "gnd": "http://d-nb.info/gnd/128862580"}],
  "date_published": "1642",
  "year": 1642,
  "place_of_publication": "Breslau",
  "publisher": "Georgius Baumann",
  "repository": "Universitätsbibliothek Breslau",
  "typeface": "Fraktur",
  "genre_dtamain": "Gebrauchsliteratur",
  "genre_dtasub": "Leichenpredigt",
  "dta_corpus_labels": ["ready", "aedit"],
  "language": "deu",
  "language_note": "(Früh-)Neuhochdeutsch",
  "license": "http://creativecommons.org/licenses/by-sa/3.0/de/",
  "license_family": "cc-by-sa-3.0",
  "num_tokens": 11515,
  "num_sentences": 842,
  "text": "Kurtze Anleitung:\nWie die jetzige boͤſe Zeit/ darinnen zwar fuͤr ſich ſelbſt/\nnichts/ alß eytel Klag/ Ach/ vnd Weh regieret; ...",
  "text_source": "layout",
  "text_normalized": "Kurze Anleitung : Wie die jetzige böse Zeit / darinnen zwar für sich selbst / nichts / als eitel Klage / Ach / und Weh regieret ; ..."
}

Dataset creation

Source data

The DTA is a reference corpus of printed German from the late 15th to the early 20th century, curated by the Berlin-Brandenburg Academy of Sciences and Humanities (BBAW). Texts were transcribed by double keying from first editions wherever possible, and are published with TEI headers, sentence and token layers and orthographic normalization from the CAB tool chain.

This dataset was built from the complete TCF dump dta_komplett_2026-02-10, which comes in two variants:

• simple/ (5,107 files): CMDI metadata and the layout-faithful historical text (<text> layer);
• full/ (5,471 files): CMDI metadata and the annotation layers (tokens, sentences, orthography).

The union of both variants gives 5,481 works. One work (ford_pitty_1633) is excluded because its TCF file is not well-formed XML.

Processing

• text comes from simple/ where available (text_source = "layout", 5,105 docs). For the 375 works only present in full/, it is reconstructed by space-joining the token layer (text_source = "tokens"), so line breaks and the original whitespace are lost there.
• text_normalized is the full/ token layer with CAB replace corrections applied (multi-token corrections replace the whole span), space-joined. It is null for the 10 works only present in simple/.
• Metadata is deliberately kept un-flattened: authors are a list of structs, all four genre schemes are separate columns, the raw license URL is kept next to a normalized license_family.
• num_tokens, num_types, num_characters, num_images and num_pages are the extent measures from the DTA header, not recomputed values.

The conversion script and the statistics script are available in the dta-experiments repository.

Statistics

All tables below were produced with dta_dataset_stats.py from the released Parquet shards.

Overview

statistic	value
Documents	5,480
Documents with normalized text (full/ layer)	5,470 (99.8%)
Documents with at least one author	3,104 (56.6%)
Distinct authors (by GND id)	1,317
Year range	1472-1987 (1 unknown)
Median year	1836
Tokens (header measure)	204,582,701
Types (header measure)	33,741,287
Characters (header measure)	1,423,995,435
Characters in text column	1,380,752,264
Characters in text_normalized column	1,370,639,606
Sentences (full/ layer)	10,376,423
Page images	761,995
Median tokens per document	10,394
Median pages (images) per document	16

Temporal distribution (50-year bins)

period	docs	sentences	tokens	characters	token share
1450–1499	15	1,899	177,258	997,389	0.1%
1500–1549	17	10,469	308,471	1,915,965	0.2%
1550–1599	98	139,035	3,572,336	23,618,803	1.7%
1600–1649	422	433,762	9,757,100	66,033,889	4.8%
1650–1699	691	1,118,361	23,338,594	159,938,400	11.4%
1700–1749	435	1,143,293	24,779,141	171,116,209	12.1%
1750–1799	583	1,566,582	30,933,941	214,031,382	15.1%
1800–1849	1,618	2,000,919	40,798,688	284,536,219	19.9%
1850–1899	1,143	3,086,868	55,792,116	393,248,560	27.3%
1900–1949	444	737,908	13,114,628	95,203,044	6.4%
1950–1999	13	137,150	2,005,488	13,321,270	1.0%
unknown	1	177	4,940	34,305	0.0%
total	5,480	10,376,423	204,582,701	1,423,995,435	100%

Text source

text_source	docs	docs %	tokens	tokens %
layout	5,105	93.2%	185,164,389	90.5%
tokens	375	6.8%	19,418,312	9.5%

Language

language	docs	docs %	tokens	tokens %
deu	5,477	99.9%	204,523,746	100.0%
lat	2	0.0%	56,714	0.0%
gml	1	0.0%	2,241	0.0%

Genre (DWDS main)

genre_dwds1main	docs	docs %	tokens	tokens %
Zeitung	2,036	37.2%	17,152,284	8.4%
Gebrauchsliteratur	1,718	31.4%	58,283,320	28.5%
Wissenschaft	953	17.4%	87,148,247	42.6%
Belletristik	773	14.1%	41,998,850	20.5%

Genre (DWDS sub, top 10)

genre_dwds1sub	docs	docs %	tokens	tokens %
unknown	2,036	37.2%	17,152,284	8.4%
Leichenpredigt	336	6.1%	3,517,225	1.7%
Zeitschrift	217	4.0%	1,697,733	0.8%
Theologie	214	3.9%	10,486,746	5.1%
Brief	204	3.7%	84,325	0.0%
Roman	198	3.6%	13,211,036	6.5%
Gesellschaft	165	3.0%	3,852,413	1.9%
Novelle	126	2.3%	3,031,211	1.5%
Lyrik	125	2.3%	5,090,960	2.5%
Gelegenheitsschrift:Tod	116	2.1%	103,280	0.1%
other (123 values)	1,743	31.8%	146,355,488	71.5%

Genre (DTA main)

genre_dtamain	docs	docs %	tokens	tokens %
unknown	3,468	63.3%	69,372,311	33.9%
Fachtext	743	13.6%	78,106,238	38.2%
Belletristik	561	10.2%	33,626,235	16.4%
Gebrauchsliteratur	536	9.8%	22,981,508	11.2%
Wissenschaftliche Abhandlungen in Form gedruckter Briefe	50	0.9%	89,564	0.0%
Abhandlungen in Zeitschriften, Sammelbänden etc.	36	0.7%	145,342	0.1%
Ankündigungen, Berichtigungen und kurze Nachrichten	29	0.5%	25,687	0.0%
Berliner Akademiereden/-schriften und andere Reden	17	0.3%	71,027	0.0%
Rezensionen	7	0.1%	8,514	0.0%
Pariser Akademiereden/-schriften	6	0.1%	16,728	0.0%
Vorworte und andere Beiträge Humboldts in Schriften anderer Autoren, Lexikonartikel	6	0.1%	20,581	0.0%
Gelegenheitsschrift	5	0.1%	49,906	0.0%
Journalismus	5	0.1%	54,829	0.0%
Albumblätter	3	0.1%	812	0.0%
Gutachten	3	0.1%	6,339	0.0%
Sonderdrucke und andere Grenzfälle zu selbständig erschienenen Schriften	2	0.0%	3,358	0.0%
Wissenschaft	2	0.0%	3,124	0.0%
Sachliteratur	1	0.0%	598	0.0%

Genre (DTA sub, top 10)

genre_dtasub	docs	docs %	tokens	tokens %
unknown	3,628	66.2%	69,896,619	34.2%
Leichenpredigt	334	6.1%	3,495,630	1.7%
Roman	163	3.0%	11,534,905	5.6%
Prosa	128	2.3%	10,291,868	5.0%
Lyrik	114	2.1%	4,731,157	2.3%
Drama	82	1.5%	2,330,599	1.1%
Recht	79	1.4%	9,743,726	4.8%
Philosophie	76	1.4%	5,426,303	2.7%
Historiographie	46	0.8%	7,221,703	3.5%
Medizin	46	0.8%	6,013,804	2.9%
other (86 values)	784	14.3%	73,896,387	36.1%

Typeface

typeface	docs	docs %	tokens	tokens %
Fraktur	4,478	81.7%	164,623,059	80.5%
Antiqua	427	7.8%	35,856,371	17.5%
Schwabacher	338	6.2%	724,934	0.4%
Handschrift	218	4.0%	1,315,690	0.6%
Jean-Paul-Fraktur	9	0.2%	1,409,611	0.7%
Current	2	0.0%	1,963	0.0%
Rotunda	2	0.0%	313,225	0.2%
Antiqua (Schreibmaschine)	1	0.0%	34,176	0.0%
Antiqua, Kursive	1	0.0%	18,883	0.0%
Antqiua	1	0.0%	3,527	0.0%
Frakur	1	0.0%	10,103	0.0%
Textur	1	0.0%	270,599	0.1%
unknown	1	0.0%	560	0.0%

Values are taken verbatim from the headers, including the two typos (Antqiua, Frakur).

DTA corpus labels (a document can carry several)

label	docs	docs %	tokens	tokens %
ready	5,480	100.0%	204,582,701	100.0%
core	1,478	27.0%	130,114,948	63.6%
china	1,149	21.0%	111,416,242	54.5%
mkhz1	625	11.4%	3,996,672	2.0%
nrhz	530	9.7%	4,290,483	2.1%
aedit	335	6.1%	3,508,098	1.7%
mts	320	5.8%	18,006,989	8.8%
lefevre	319	5.8%	476,390	0.2%
dtae	244	4.5%	21,475,420	10.5%
mkhz2	222	4.1%	2,169,371	1.1%
correspondent	204	3.7%	925,755	0.5%
tevo	202	3.7%	5,682,395	2.8%
sanders-briefe	190	3.5%	61,890	0.0%
hab	184	3.4%	10,256,959	5.0%
augsburgerallgemeine	181	3.3%	2,761,778	1.3%
avh	172	3.1%	443,668	0.2%
wikisource	169	3.1%	5,787,543	2.8%
sbb_funeralschriften	112	2.0%	77,981	0.0%
tdef	109	2.0%	940,772	0.5%
novellenschatz	87	1.6%	1,643,336	0.8%
blumenbach	33	0.6%	2,688,412	1.3%
epoetics	23	0.4%	3,074,730	1.5%
frauenstudium	22	0.4%	134,198	0.1%
psyleko	19	0.3%	2,075,500	1.0%
ntsm	11	0.2%	32,139	0.0%
avhkv	9	0.2%	636,539	0.3%
briefjeanpaul	9	0.2%	1,409,611	0.7%
gutenberg_org	8	0.1%	325,940	0.2%
gutzkow	7	0.1%	478,246	0.2%
dwds1	5	0.1%	465,688	0.2%
gei	2	0.0%	57,474	0.0%
gutenberg_de	2	0.0%	116,184	0.1%
gwb	2	0.0%	80,666	0.0%
urmel	2	0.0%	1,068	0.0%
zbk	2	0.0%	48,127	0.0%
greflinger	1	0.0%	9,079	0.0%
grenzboten	1	0.0%	116,208	0.1%

The core label marks the curated DTA core corpus (1,478 works, 63.6% of all tokens), which is the balanced selection by genre and period. The other labels denote extension corpora (DTAE) contributed by partner projects.

Place of publication (top 10)

place	docs	docs %	tokens	tokens %
Berlin	604	11.0%	30,616,946	15.0%
Köln	532	9.7%	4,682,997	2.3%
Leipzig	425	7.8%	41,950,384	20.5%
Hamburg	330	6.0%	5,514,413	2.7%
Augsburg	256	4.7%	4,581,351	2.2%
Altstrelitz	172	3.1%	57,702	0.0%
unknown	160	2.9%	980,933	0.5%
München	153	2.8%	3,434,662	1.7%
Frankfurt (Main)	141	2.6%	11,430,873	5.6%
Königsberg	133	2.4%	926,843	0.5%
other (246 values)	2,574	47.0%	100,405,597	49.1%

Most frequent authors (top 10)

author	GND	docs	tokens
Humboldt, Alexander von	http://d-nb.info/gnd/118554700	183	1,504,828
Sanders, Daniel	http://d-nb.info/gnd/119242044	173	69,412
Dach, Simon	http://d-nb.info/gnd/11852321X	112	77,981
N. N.		51	1,882,041
Blumenbach, Johann Friedrich	http://d-nb.info/gnd/116208503	35	2,814,553
Sattler, Basilius	http://d-nb.info/gnd/116974788	26	352,365
Goethe, Johann Wolfgang von	http://d-nb.info/gnd/118540238	24	1,064,748
Herder, Johann Gottfried von	http://d-nb.info/gnd/118549553	23	690,276
Grimm, Jacob	http://d-nb.info/gnd/118542257	17	2,441,309
Fontane, Theodor	http://d-nb.info/gnd/118534262	16	1,262,140

Licensing

The DTA publishes every work under its own license, and this dataset keeps that information per row in the license (raw URL from the header) and license_family (normalized) columns. There is no single license for the whole dataset. Note that only cc0-1.0 is a public domain dedication; all cc-by-* licenses carry binding conditions (attribution, share-alike, or non-commercial use), and about 28% of the documents are restricted to non-commercial use (cc-by-nc-3.0, cc-by-nc-sa-4.0, noc-nc-1.0, out-of-copyright-nc).

license family	docs	docs %	tokens	tokens %	license URLs
cc-by-sa-4.0	2,644	48.2%	139,433,878	68.2%	http://creativecommons.org/licenses/by-sa/4.0/ http://creativecommons.org/licenses/by-sa/4.0/deed.de/ https://creativecommons.org/licenses/by-sa/4.0/ https://creativecommons.org/licenses/by-sa/4.0/deed.de
cc-by-nc-3.0	1,559	28.4%	20,418,259	10.0%	http://creativecommons.org/licenses/by-nc/3.0/ http://creativecommons.org/licenses/by-nc/3.0/de http://creativecommons.org/licenses/by-nc/3.0/de/
cc-by-sa-3.0	623	11.4%	15,287,129	7.5%	http://creativecommons.org/licenses/by-sa/3.0/ http://creativecommons.org/licenses/by-sa/3.0/de http://creativecommons.org/licenses/by-sa/3.0/de/ https://creativecommons.org/licenses/by-sa/3.0/ https://creativecommons.org/licenses/by-sa/3.0/de/
cc0-1.0	370	6.8%	16,938,595	8.3%	http://creativecommons.org/publicdomain/zero/1.0/
cc-by-4.0	215	3.9%	8,189,539	4.0%	http://creativecommons.org/licenses/by/4.0/ http://creativecommons.org/licenses/by/4.0/deed.de https://creativecommons.org/licenses/by/4.0/ https://creativecommons.org/licenses/by/4.0/deed.de
cc-by-sa-2.0	51	0.9%	3,058,054	1.5%	http://creativecommons.org/licenses/by-sa/2.0/de http://creativecommons.org/licenses/by-sa/2.0/de/
gutenberg	9	0.2%	344,316	0.2%	https://web.archive.org/web/20180927123034/http://www.gutenberg.org/wiki/Gutenberg:The_Project_Gutenberg_License
cc-by-3.0	3	0.1%	571,346	0.3%	http://creativecommons.org/licenses/by/3.0/de/
cc-by-nc-sa-4.0	2	0.0%	19,672	0.0%	https://creativecommons.org/licenses/by-nc-sa/4.0/ https://creativecommons.org/licenses/by-nc-sa/4.0/deed.de
mdz-copyright	1	0.0%	24,964	0.0%	http://mdz.bib-bvb.de/copyright.htm
noc-nc-1.0	1	0.0%	125,859	0.1%	http://rightsstatements.org/vocab/NoC-NC/1.0/
nug-kkn	1	0.0%	55,702	0.0%	http://www.deutsche-digitale-bibliothek.de/lizenzen/nug-kkn/
out-of-copyright-nc	1	0.0%	115,388	0.1%	https://www.europeana.eu/portal/de/rights/out-of-copyright-non-commercial.html
total	5,480	100%	204,582,701	100%

To select a subset that fits your use case:

ds_sa = ds.filter(lambda d: d["license_family"] in {"cc0-1.0", "cc-by-4.0", "cc-by-3.0", "cc-by-sa-4.0", "cc-by-sa-3.0", "cc-by-sa-2.0"})

Considerations for using the data

• Coverage is uneven. Roughly 60% of the tokens date from 1750 to 1899; the 15th and 16th centuries are thinly covered. Newspapers (Zeitung) dominate the document count but make up only 8% of the tokens, and a single project (Alexander von Humboldt's writings) contributes 183 short documents.
• Historical content. The texts reflect the views, language and stereotypes of their time, including religious, political and colonial-era writing. They are provided for research on historical German and should not be taken as statements of fact or acceptable opinion.
• Normalization is automatic. text_normalized comes from the CAB tool chain and is not manually corrected. Expect residual errors, especially for very early texts, Latin passages, and proper names. The normalized text is also whitespace-tokenized (punctuation is separated by spaces), whereas text keeps the original layout.
• text is not uniform. 375 documents have their historical text reconstructed from the token layer and lack line breaks; check text_source if layout matters to you.
• Metadata quirks. Genre fields are null for a large share of the works in the DTA scheme (use the DWDS scheme for complete main-genre coverage), typeface strings contain a few typos, and year is parsed heuristically from the free-text date_published.

Citation

Please cite the DTA when using this dataset:

@misc{dta2026,
  author       = {{Berlin-Brandenburgische Akademie der Wissenschaften}},
  title        = {Deutsches Textarchiv. Grundlage für ein Referenzkorpus der neuhochdeutschen Sprache},
  year         = {2026},
  address      = {Berlin},
  howpublished = {Herausgegeben von der Berlin-Brandenburgischen Akademie der Wissenschaften},
  url          = {https://www.deutschestextarchiv.de/}
}

Acknowledgements

A big thank you to the whole team of the Deutsches Textarchiv at the Berlin-Brandenburg Academy of Sciences and Humanities! Also many thanks to all the partner projects, libraries and archives that contributed texts to the DTA extension corpora.

This dataset only converts an official dump into Parquet for convenient use with the Hugging Face ecosystem.

AI disclosure

This dataset card was drafted with Claude Fable 5 (claude-fable-5) based on the statistics produced by dta_dataset_stats.py, and reviewed by the dataset author. The conversion and statistics scripts carry their own AI disclosure blocks, following the rules in the ai-disclosure repository.