English Models

• PRE: tools that need to be run prior to the model.

• DATA: the dataset used to build the model.

• EVAL: evaluation on the dataset that the model is trained on.

• BM: the standard benchmark evaluation for the task.

• tokseg: any tokenizer with sentence segmentation.

• inmixed: any dataset in the mixed corpus.

Tokenization

Tokenization Models

Model ID
elit_tok_space_un
elit_tok_lexrule_en

Morphological Analysis

Model ID	PRE
elit_morph_idprule_en	tokseg

Part-of-Speech Tagging

Model ID	PRE	DATA	EVAL	BM
elit_pos_flair_en_mixed	tokseg	Mixed	97.80%	97.72%

• EVAL: accuracy.

• BM: accuracy on the Wall Street Journal portion of the Penn Treebank using the standard split (trn: 0-18; dev: 19-21; tst: 22-24).

Named Entity Recognition

Model ID	PRE	DATA	EVAL	BM
elit_ner_flair_en_ontonotes	tokseg	OntoNotes	88.75%	92.74%

• EVAL: F1-score.

• BM: F1-score on the English dataset distributed by the CoNLL 2003 shared task.

Dependency Parsing

Model ID	PRE	DATA	EVAL	BM
elit_dep_biaffine_en_mixed	tokseg	Mixed	92.26/91.03	96.08/95.02

• EVAL: UAS (unlabeled attachment score) / LAS (labeled attachment score).

• BM: UAS/LAS on the Wall Street Journal portion of the Penn Treebank using the standard split (trn: 2-21; dev: 22, 24; tst: 23) and the Stanford typed dependencies.

Semantic Dependency Parsing

Model ID	PRE	DATA	EVAL	BM
elit_sdp_biaffine_en_mixed	tokseg	Mixed	?	90.68/85.34

• EVAL: Labeled F1 score.

• BM: Average labeled F1 scores on the in-domain and out-of-domain test sets distributed by the SemEval 2015 shared task.