Parsan — Hebrew morphosyntax

Paste Hebrew text and get Universal Dependencies: morphological segmentation, POS, features, lemmas, and a dependency parse. The first run downloads the models, so the first parse takes a minute.

Parsan: code · models. Lemmas come from the published lookup bank (271K entries) with the neural edit-script head as backoff, the same configuration as the paper. To hand-edit a parse, load the CoNLL-U into UD Annotatrix (Tyers, Washington et al.), a full Universal Dependencies treebank editor.