id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_pu54zjotk5ed3ol4ajhrlzpfju	Tobias Schnabel	FLORS: Fast and Simple Domain Adaptation for Part-of-Speech Tagging	2014	12	.pdf	application/pdf	8156	832	73	This representation consists of distributional features, suffixes and word shapes of v and its local FLORS predicts unseen tags of known words better than prior work on DA for POS. FLORS uses representations computed from unlabeled text, representations of unknown words are For a word w occurring as token vi in a sentence, we build a feature vector for a local window Table 2: Tagging accuracy of four baselines and FLORS on the dev sets. (lines 1–4), basic FLORS setup (lines 5–6), effect of omitting one of the three feature types if the word to be tagged designed for tagging in-domain data and use feature In contrast to standard approaches to POS tagging, the FLORS basic representation does not contain vocabulary indices. Table 3: Tagging accuracy of four baselines and FLORS on the test sets. Table 7: Tagging accuracy of different word representations on the dev sets.	./cache/work_pu54zjotk5ed3ol4ajhrlzpfju.pdf	./txt/work_pu54zjotk5ed3ol4ajhrlzpfju.txt