id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_kewvexetpzdd5axo24uef7rq64	Jonathan Berant	Imitation Learning of Agenda-based Semantic Parsers	2015	14	.pdf	application/pdf	9293	1048	71	Imitation Learning of Agenda-based Semantic Parsers In this paper, we combine ideas from imitation learning and agendabased parsing to train a semantic parser that Figure 2: An example semantic parse, or derivation, for the beam search, where the number of parses (see Figure 2) for each chart cell (e.g., (SET,3:5)) is capped Specifically, we cast agenda-based semantic parsing as a Markov decision process, where that learns to choose good parsing actions, training from question-answer pairs only; Second, a lazy Figure 3: A semantic function (we show JOIN, LEX and INTERSECT) takes one or two child derivations and returns a set of controls the order in which derivations are constructed using an agenda Q, which contains a set of Most work on agenda-based parsing generally assumed that the scoring function s is Table 4: Accuracy, number of featurized derivations, and parsing time for both the training set and development set when	./cache/work_kewvexetpzdd5axo24uef7rq64.pdf	./txt/work_kewvexetpzdd5axo24uef7rq64.txt