noamor/shoshan-data

Shoshan — Hebrew lemmatization data

One content lemma per surface token, in context. Used to train and evaluate the Shoshan lemmatizer.

file	split	rows
train.csv / dev.csv / test.csv	in-domain (Knesset + Wikipedia, IAHLT UD)	191k / 11k / 11k
ood.csv (+ ood_Bagatz.csv, ood_GeekTime.csv, ood_Dicta.csv)	out-of-domain benchmark, 100 sentences/domain	~5k
oov.csv	unseen-lemma tail	100

Columns: form, lemma, pos, sentence, source, sent_id.

Load with the package:

from shoshan import data
df = data.load("ood")      # -> pandas DataFrame, cached after first download

License CC BY 4.0. Derived from the IAHLT Hebrew UD treebank (Knesset + Wikipedia) and public Hebrew lexicons; see the code repo's docs/DATA_STATEMENT.md for full provenance. We thank Avner Algom and the IAHLT.