Cross-Register Projection for Headline Part of Speech Tagging

<p>POSH: The POS-tagged HeadlIne corpus was created for the paper &ldquo;Cross-Register Projection for Headline Part of Speech Tagging&rdquo; published in EMNLP 2021.&nbsp; This dataset contains headlines with gold annotated POS tags.</p> <p>The <em>GSCh</em> evaluation set is here: GSCh/gsc-headline-gold.test.conllu</p> <p>The smaller evaluation set of GSC headlines sampled uniformly at random (described in section 2.3) is here: gold_unconstrained_headlines/unifrand_gsc.test.conllu</p> <p>The POS-tagged NYT headlines described in section 2.3 are not shared directly as this text was drawn from the New York Times Annotated Corpus (LDC2008T19), and subject to license constraints.&nbsp; However, if you have access to and have untarred LDC2008T19, you can recover this evaluation set with:</p> <p>&nbsp;&nbsp;&nbsp; TAG_PATH=&quot;./unifrand_onlynyt.tags.json&quot;&nbsp; # mapping from NYT headline span to gold POS tag</p> <p>&nbsp;&nbsp;&nbsp; python build_gold_nyt_headlines.py --nyt_dir /PATH/TO/ANNOTATED/NYT/CORPUS/ --tag_path ${TAG_PATH} --num_proc 4</p> <p>Increase the argument to --num_procs to process more shards from the NYT corpus in parallel and reduce build time.</p> <p>Under GSCproj we also share the <em>GSCproj</em> folds which we used to train and validate our models.&nbsp; These are not gold POS tags, and are shared purely for reproducibility sake.</p>