Features

Features

Feature names in pkl	Description (Szarvas et al. 2012)
lemma_-2 lemma_-1 lemma_0 lemma_1 lemma_2	Stems of the current token and the tokens in a window of size 2 (two tokens to the left and right).
pattern_prefix pattern_-1 pattern_0 pattern_1	Surface patterns of the current token and the tokens in a window of size one (one token to the left and right). This feature describes the capitalization and other orthographic features.
prefix_3 prefix_4 prefix_5 suffix_3 suffix_4 suffix_5	Prefixes and suffixes of the current token, 3-5 characters long.
pos_-2 pos_-1 pos_0 pos_1 pos_2	Part-of-speech (POS) tags of the current token and the tokens in a window of size 2.
chunk_-2 chunk_-1 chunk_0 chunk_1 chunk_2	Syntactic chunks of the current token and the tokens in a window of size 2.

* Szarvas et al. (2012) use an additional type of features: concatenated stem, POS, and chunk labels. This feature was not used in the current system.