Integrating Multiple Dependency Corpora for Inducing Wide-Coverage Japanese CCG Resources

Abstract

This paper describes a method of inducing wide-coverage CCG resources for Japanese. While deep parsers with corpusinduced grammars have been emerging for some languages, those for Japanese have not been widely studied, mainly because most Japanese syntactic resources are dependency-based. Our method first integrates multiple dependency-based corpora into phrase structure trees and then converts the trees into CCG derivations. The method is empirically evaluated in terms of the coverage of the obtained lexicon and the accuracy of parsing.

Improving Dependency Parsers using Combinatory Categorial Grammar

“…Conversely, Ambati et al (2013) showed that a Hindi dependency parser (Malt) could be improved by using CCG categories. Using an algorithm similar to Cakici (2005) and Uematsu et al (2013), they first created a Hindi CCGbank from a Hindi dependency treebank and built a supertagger. They provided CCG categories from a supertagger as features to Malt and obtained overall improvements of 0.3% and 0.4% in unlabelled and labelled attachment scores respectively.…”

Hindi CCGbank: A CCG treebank from the Hindi dependency treebank

“…There has also been work on extracting CCG lexicons (Cakici 2005) and CCGbanks (Bos et al 2009;Uematsu et al 2013Uematsu et al , 2015 from dependency treebanks. Bos et al (2009) created an Italian CCGbank from the Turin University Treebank (TUT), 1 an Italian dependency treebank.…”

“…They first converted dependency trees into phrase structure trees and then applying an algorithm similar to Hockenmaier and Steedman (2007) extracted the CCG derivations. Using different dependency resources available for Japanese like the Kyoto corpus (Kawahara et al 2002) and the NAIST text corpus (Iida et al 2007), Uematsu et al (2013) developed a CCGbank for Japanese. They first integrated the dependency resources into phrase structure trees and then converted them into CCG derivations.…”

Jigg: A Framework for an Easy Natural Language Processing Pipeline

“…Jigg thus supports many Japanese processing tools such as MeCab (Kudo et al, 2004), a famous morphological analyzer, as well as a Japanese CCG parser based on the Japanese CCGBank (Uematsu et al, 2013). For English, currently the core tool is Stanford CoreNLP.…”