id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_v2l3qcuonvdkxoeb5oplbydezm	Yufan Guo	Unsupervised Declarative Knowledge Induction for Constraint-Based Learning of Information Structure in Scientific Documents	2015	14	.pdf	application/pdf	8276	831	61	documents, induces constraints from this information and maps sentences to their dominant information structure categories through (2013a) recently applied the Generalized Expectation (GE) criterion (Mann and McCallum, 2007) to information structure analysis using This approach, however, requires human supervision in several forms including task specific constraint templates (see Section 2). for a GE model and a bias term for a graph clustering objective, such that the resulting models assign each of the input sentences to one information Information Structure Learning with Declarative Knowledge Recently, Reichart and Korhonen We therefore constructed two types of topic models: section-specific and article-level models, reasoning that some distinctions apply globally at the Table 6: ROUGE scores of zone (TopicGE, TopicGC, ExpertGE or gold standard) and Discussion section based summaries. Table 7: Topics and key features extracted by the article-level model (including modal, tense and voice marked in constraint-based modeling of the information structure analysis of scientific documents.	./cache/work_v2l3qcuonvdkxoeb5oplbydezm.pdf	./txt/work_v2l3qcuonvdkxoeb5oplbydezm.txt