yuanxin112/wiktionary-morph

Wiktionary Morphology Dataset (German / English / French)

A cleaned, morphologically-annotated dictionary of content words built from Wiktionary, intended as raw material for designing morphology-aware tokenizer benchmarks. There is one record per word (parts of speech merged): each record carries the word's POS list, frequency signals, a morphological-structure label, the full inflectional paradigm, definitions with examples, and translations. The schema is flat (all fields at the top level, like the underlying wiktextract data).

Config	Words	Source edition
de	92,088	German Wiktionary (de.wiktionary.org)
en	97,303	English Wiktionary (en.wiktionary.org)
fr	69,476	French Wiktionary (fr.wiktionary.org)

from datasets import load_dataset
de = load_dataset("yuanxin112/wiktionary-morph", "de", split="train")
en = load_dataset("yuanxin112/wiktionary-morph", "en", split="train")
fr = load_dataset("yuanxin112/wiktionary-morph", "fr", split="train")

Data fields

Field	Type	Description
word	string	The word (headword). One record per word.
pos	list[str]	All parts of speech the word has — open class (noun, verb, adj, adv) and closed class (prep, conj, pron, det, num, intj, particle, …); e.g. ["verb","noun","adj"] for run.
lang_code	string	de, en, or fr.
freq	int	Frequency in a BabyLM training corpus, summed over the whole inflectional paradigm across all POS (the word and all its inflected forms). Corpus size: English ≈ 10M words; German and French ≈ 100M words (French = BabyLM-community/babylm-fra). Kept even when 0.
citation_freq	int	Frequency of the citation/headword form only.
zipf	float	wordfreq Zipf frequency from large general corpora (0 = unattested, ~3 = rare, ~5 = common). External rarity signal, independent of the training corpus and comparable across languages.
morph_type	string	The word's own structure: derivation, compound, or simple (see below).
topics	list[str]	Domain/topic labels aggregated over senses (e.g. navy, botany).
hyphenation	list[str]	Syllable split (e.g. ["be","ar","bei","ten"]). null for French (not provided by French Wiktionary).
etymology_text	string	Raw etymology text from Wiktionary.
decomposition	struct	English config only. {base, affixes, parts} — base + affixes for derivations, parts for compounds — taken only from Wiktionary's explicit affix/compound templates (prefix/suffix/affix/compound/…); null where none. Absent in de/fr (use a dedicated resource such as UniMorph).
inflection_forms	list	Full paradigm; each {form, pos, tags, pronouns} (pos = which part of speech the form belongs to).
derived	list[str]	Word-formation family: words derived from this one.
related	list[str]	Related words (same root, not direct derivations).
senses	list	Each {gloss, pos, tags, examples, sense_index} (pos = the part of speech of that sense).
synonyms	list[str]	Synonyms (German "Sinnverwandte Wörter" are merged in here).
antonyms	list[str]	Antonyms.
translations	list	Each {lang_code, lang, word, sense_index, roman}.

One record per word. A word that is several parts of speech (e.g. run, blackboard, manger) is a single record; pos lists every part of speech, and each sense and inflected form is tagged with the pos it belongs to.

morph_type

Value	Meaning	Example
derivation	Formed by affixation	bearbeiten = be- + arbeiten; happiness = happy + -ness; aborder = a- + border
compound	Formed by combining stems	Unterseeboot = Untersee + Boot; blackboard = black + board; caméscope = caméra + magnétoscope
simple	Monomorphemic / no affixation or compounding in its etymology	Haus, dog, travailler

morph_type reflects Wiktionary's etymological analysis. A word borrowed as a whole (e.g. editor < Latin *editor*) is labelled simple, even if it looks like *edit* + *-or*. When POS readings disagree, the most structured label wins (compound > derivation > simple).

Frequency signals — why several

• freq answers "how often does this word's paradigm occur in our training data?" — used to mark seen vs. unseen items for generalization splits. freq == 0 is kept on purpose: unseen derived/compound words are the most valuable generalization test cases.
• citation_freq is the narrow headword-form count.
• zipf answers "is this a word people actually use, anywhere?" — based on Wikipedia, subtitles, news, books and web text combined, so it is robust to gaps in the training corpus.

How the data was filtered

Starting from the full Wiktionary edition, an entry is dropped if it matches any of:

1. Not the target language (lang_code).
2. Part of speech is not a word category — removes proper names (pos = name), multi-word phrases, abbreviations, symbols and bound affixes (prefix/suffix). Open- and closed-class words are kept.
3. Surface form contains non-letters / whitespace (non-words, multi-word strings).
4. Inflected-form entry (tagged form-of, or a French/German/English inflection gloss such as "Féminin pluriel de…", "plural of…") — inflection is instead embedded inside inflection_forms.
5. Alternative spelling / variant (alt_of, or a "variant/alternative spelling of…" gloss).
6. Loanword / transliteration etymology and low frequency (freq < 5) — naturalized loans (e.g. Computer, Restaurant) are kept.
7. Rare / dialectal / obsolete — only if all senses are tagged rare/dialectal/regional/uncommon/obsolete/archaic/dated/nonstandard.
8. Science / bio-chemical — only if all senses carry a chemistry/biology/medicine/botany/zoology/anatomy/… topic.
9. Never attested in general corpora — wordfreq Zipf ≤ 0 (removes ultra-rare/technical junk such as alectinib, hexacosanol).

Proper names are handled implicitly: Wiktionary stores them as separate pos = name entries, so dropping those still keeps homographic common words (e.g. the eel Aal).

Provenance

• Built from the kaikki.org wiktextract machine-readable dumps of Wiktionary (German edition 2026-05; English-language extract 2026-06; French edition 2026-06).
• zipf values from the wordfreq library.
• freq / citation_freq from a BabyLM training corpus per language — English ≈ 10M words, German and French ≈ 100M words (French = BabyLM-community/babylm-fra).

Limitations

• decomposition is provided for English only and comes solely from Wiktionary's explicit affix/compound templates (high precision; present for ~55% of English derivations/compounds, null otherwise). German and French omit decomposition; use etymology_text or an external resource (e.g. UniMorph).
• hyphenation is unavailable for French.
• Coverage and labels are only as good as Wiktionary; rare/technical entries may be imperfectly tagged.
• The dataset is a snapshot of the dump dates above, not live Wiktionary.

License

Derived from Wiktionary, licensed under CC BY-SA 4.0 (and GFDL). Released under CC BY-SA 4.0; attribute Wiktionary and its contributors.