A* CCG Parsing with a Supertag and Dependency Factored Model

Abstract

We propose a new A* CCG parsing model in which the probability of a tree is decomposed into factors of CCG categories and its syntactic dependencies both defined on bi-directional LSTMs. Our factored model allows the precomputation of all probabilities and runs very efficiently, while modeling sentence structures explicitly via dependencies. Our model achieves the stateof-the-art results on English and Japanese CCG parsing. 1

Acquisition of Phrase Correspondences Using Natural Deduction Proofs

“…For the natural deduction proofs, we used ccg2lambda (Martínez-Gómez et al, 2016) 5 , a higher-order automatic inference system, which converts CCG derivation trees into semantic representations and conducts natural deduction proofs automatically. We parsed the tokenized sentences of the premises and hypotheses using three widecoverage CCG parsers: C&C (Clark and Curran, 2007), EasyCCG (Lewis and Steedman, 2014), and depccg (Yoshikawa et al, 2017). CCG derivation trees (parses) were converted into logical semantic representations based on Neo-Davidsonian event semantics (Section 3.1).…”

Automatic Generation of High Quality CCGbanks for Parser Domain Adaptation

“…Additionally, we manually construct experimental data for parsing (4) math problems (Seo et al, 2015), for which the importance of domain adaptation is previously demonstrated . We observe huge additive gains in the performance of the depccg parser (Yoshikawa et al, 2017), by combining contextualized word embeddings and our domain adaptation method: in terms of unlabeled F1 scores, 90.68% to 95.63% on speech conversation, and 88.49% to 95.83% on math problems, respectively. 2 Figure 2: Example CCG derivation tree for phrase cats that Kyle wants to see.…”

“…We evaluate our method in terms of performance gain obtained by fine-tuning an off-the-shelf CCG parser depccg (Yoshikawa et al, 2017), on a variety of CCGbanks obtained by converting existing dependency resources using the method. In short, the method of depccg is equivalent to omitting the dependence on a dependency tree z from P (y|x, z) of our converter model, and running an A* parsing-based decoder on p tag|dep calculated on h 1 , ..., h N = Ω(e x 1 , ..., e x N ), as in our method.…”

“…In the plain depccg, the word representation e x i is a concatenation of GloVe 7 vectors and vector representations of affixes. As in the previous work, the parser is trained on both the English CCGbank (Hockenmaier and Steedman, 2007) and the tri-training dataset by Yoshikawa et al (2017). In this work, on top of that, we include as a baseline a setting where the affix vectors 7 https://nlp.stanford.edu/projects/ glove/ ), 8 which we find marks the current best scores in the English CCGbank parsing ( Table 1).…”

“…The recent advancement of Combinatory Categorial Grammar (CCG; Steedman (2000)) parsing Yoshikawa et al, 2017), combined with formal semantics, has enabled high-performing natural language inference systems (Abzianidze, 2017;Martínez-Gómez et al, 2017). We are interested in transferring the success to a range of applications, e.g., inference systems on scientific papers and speech conversation.…”