id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_hea3gijzhnc2rl4sxgyj6pvghm	Annie Louis	What Makes Writing Great? First Experiments on Article Quality Prediction in the Science Journalism Domain	2013	12	.pdf	application/pdf	8028	639	65	For example, content-based recommendation systems standardly represent user interest using frequent words from articles in a user's history readable, well-written and topically interesting articles, giving an accuracy of 84% (Section 5). We collect a larger set of visual words from a corpus of tagged images from the ESP game (von Ahn We compute the precision of each topic as the proportion of these 200 words that match the MRC list Topic-based features: We also compute what proportion of the words we identify as visual matches We also compute a greedy cover set of topics for the visual words in the article. These features capture the mix of visual words from different topics. GOOD articles contain more visual words overall as a feature (SURP) and also this value normalized by total number of word tokens in the article We normalize each feature by the total words in the article.	./cache/work_hea3gijzhnc2rl4sxgyj6pvghm.pdf	./txt/work_hea3gijzhnc2rl4sxgyj6pvghm.txt