---
pretty_name: "Rafb\xF3kavefurinn"
language:
- is
license: other
license_name: Icelandic Copyright Law
size_categories:
- n<1K
task_categories:
- text-generation
- fill-mask
task_ids:
- language-modeling
source_datasets:
- original
domains:
- Books
---
# Dataset Card for Rafbókavefurinn
Out-of-copyright Icelandic books, from the medieval sagas to the early twentieth century.
Three kinds of material sit side by side:
- the legendary *fornaldarsögur Norðurlanda*, the Íslendingasögur and the large historical compilations (*Heimskringla*, *Landnámabók*);
- a New Testament translation;
- and printed books from roughly the 1850s to the 1940s — novels, short stories, folk tales, a lecture on women's rights, and translations of Homer, Dumas, Verne, Twain, Andersen, Poe and *Þúsund og ein nótt*.
The medieval texts appear in two different orthographic traditions. The *fornaldarsögur* keep classical Old Norse spelling, while the Íslendingasögur are given in modernised spelling, so the same language is represented at both ends of the normalisation question.
## Dataset Description
- **Number of samples**: 129
- **Number of tokens (Llama 3)**: 11.27M
- **Average document length in tokens (min, max)**: 87.40K (1.63K, 616.99K)
## Dataset Structure
An example from the dataset looks as follows.
```py
{
"id": "rafbokavefur_22",
"text": "SAGAN AF\n\nHELJARSLÓÐARORRUSTU\n\neftir Benedikt Gröndal\n\n1. kapítuli\n\nSvo segja vitrir menn, að sá hlu[...]",
"source": "rafbokavefur",
"added": "2026-08-17",
"created": "1861-01-01, 1861-12-31",
"token_count": 62228
}
```
### Data Fields
An entry in the dataset consists of the following fields:
- `id` (`str`): A unique identifier for each document.
- `text` (`str`): The content of the document.
- `source` (`str`): The source of the document (see [Source Data](#source-data)).
- `added` (`str`): The date when the document was added to this collection.
- `created` (`str`): The date range when the document was originally created.
- `token_count` (`int`): The number of tokens in the sample computed using the Llama 3 tokenizer.
The numeric part of the `id` is the book's own post number on Rafbókavefurinn, so each document
can be traced back to the page it came from.
`created` needs care. For most printed books it is the year of the edition the text was set from,
which the books state themselves. The medieval texts have no such year and share one broad range
covering the period in which they were composed. Two printed books state no year either, and fall
back to a range spanning the whole period the printed part of the collection covers.
### Dataset Statistics
### Annotation Overview
Each document comes with annotations describing its content, such as content quality, information density, and educational value.
Each bar shows the share of documents at each level of one annotation, from worst (light) to best (dark).
To learn more about the annotations and how to use them, see the [annotations section](https://huggingface.co/datasets/danish-foundation-models/icelandic-dynaword#annotations) of the main readme.
## Additional Information
### Data Collection and Processing
The catalogue is enumerated through the site's own WordPress API and each book is taken from its
EPUB. Re-running the build picks up titles added since the last one.
Material that is not part of the book itself is removed, such as the imprint page, notes about
the digitisation, and the running title that repeats at the top of every chapter.
A few of the posts from Rafbókavefurinn are left out for reasons of tidiness:
- Three of the site's posts are volumes one to three of the H.C. Andersen tales and all link to a
single EPUB containing all three, so that book is included once rather than three times.
- The New Testament is also published as one post per book of the testament, none of which carries an EPUB; the complete testament is a single EPUB elsewhere in the catalogue and that is what is used.
**Most of the catalogue carries the site's *Opið efni* ("open content") tag, and those books are
taken on that basis.** The rest either states restrictive terms or no terms at all. The books with
no terms were checked individually according to the rules listed in [Licence Information](#license-information). That check removed:
- *Still in copyright.* *Klækir* and *Ást er þjófnaður*, by living authors; *Af hverju ég er ekki
kristinn*, whose Icelandic translation dates from 2006; and *Andlegt sjálfstæði*, part of which
was translated for a 2008 edition by a translator who is alive.
- *Cannot be established.* *Svipurinn hennar* names its translator, Jóhannes Vigfússon, but no
death date for him could be found. *Kvenfrelsiskonur* is attributed only to "St. Daníelsson",
who could not be identified confidently enough to date.
- *A modern editor's own writing alongside the free text.* *Sturlunga saga* carries a chronological
table by dr. Jón Jóhannesson (d. 1957), and the *Eddukvæði* close with a note quoted from Guðni
Jónsson's *Eddulyklar* (he died in 1974).
### Limitations
40 books duplicate [saga](../saga/saga.md), 14 duplicate
[wikisource](../wikisource/wikisource.md) and four duplicate
[icepahc](../icepahc/icepahc.md), some of them more than one.
The Íslendingasögur and the large compilations are the same works `saga` carries, and the
nineteenth-century novels are largely the same ones Wikisource volunteers have transcribed. The
remaining 81 books, about 63% of the tokens, are not substantially present anywhere else in the
collection.
Orthography varies widely, both between the Old Norse and modernised medieval texts and across a
century of printed books.
About half the printed books are translations, and they are the longer ones — roughly 60% of that
part of the subset by volume — so a majority of the non-medieval text is translated Icelandic
rather than originally composed.
Footnote markers and footnote text from the printed originals are
kept. Illustration captions are kept where they are text; the images themselves are not.
### Source Data
[Rafbókavefurinn](https://rafbokavefur.is) is a volunteer project that digitises Icelandic
books and publishes them as EPUBs. It was founded in 2011 by Óli Gneisti Sóleyjarson, and its
texts are proofread by volunteers through a distributed proofreading site.
The books name the sources they were set from. The medieval texts come from Netútgáfan
(`snerpa.is`), and the printed books were keyed from named editions, which each book states on its
publication-details page.
## License Information
Two bodies of law apply. The books are **Icelandic** works, and the dataset is published in the
**United States**, so every title has to be out of copyright under both. A few were printed abroad,
in Winnipeg, which changes neither test.
**Iceland** — [Höfundalög nr. 73/1972](https://www.althingi.is/lagas/nuna/1972073.html):
- Article 43: a work is protected until 70 years after the new year following the author's death.
- Article 44: a work published without naming its author is protected for 70 years from publication.
- Article 5: a translation has its own copyright, running from the **translator's** death. An
ancient original therefore says nothing about whether an Icelandic translation of it is free.
**United States** — anything published before 1931, whether inside the US or outside it, is in the
[public domain](https://guides.library.cornell.edu/copyright/publicdomain). After that the two
diverge: a foreign work published between 1931 and 1977 falls under
[17 U.S.C. § 104A](https://www.law.cornell.edu/uscode/text/17/104A), which restored US copyright in
1996 for works still protected in their home country on that date and gives them 95 years from
publication, while a US work of the same period turns on whether its copyright was renewed.
**Rafbókavefurinn** — most of the catalogue is tagged *Opið efni* ("open content"):
*"Öllum er frjálst að nota rafbækur sem eru merktar Opið efni […] Það má bæði dreifa og breyta
þessu efni að eigin vild"* — anyone may use them, and may distribute and alter them, with the
origin acknowledged. 117 of the books here carry that tag. The remaining 12 state no terms, and
are included because they are out of copyright by age under the articles above; a few books that
assert copyright or carry a permission running only to Rafbókavefurinn are left out.
The books that Rafbókavefurinn obtained from [Netútgáfan](https://www.snerpa.is/net/index.html) (96 of the books included in this dataset) further state in their front matter that they are free to build on, provided the origin is always stated clearly, and that a book shared online must *"vísa til Rafbókavefsins með tengli"* — link back to Rafbókavefurinn.
This subset reproduces the books in full, so anyone redistributing it should carry both that link and the reference to Netútgáfan forward.
Rafbókavefurinn separately *recommends* a CC BY-NC-SA licence for derivative works. That is a
recommendation rather than a condition.