Automatic Acquisition of Lexical-Functional Grammar Resources from a Japanese Dependency Corpus

Automatic Acquisition of Lexical-Functional Grammar Resources from a Japanese Dependency Corpus* Masanori Oya* and Josef van Genabith* *National Centre for Language Technology and School of Computing, Dublin City University, Dublin, Ireland {moya, josef}@computing.dcu.ie Abstract. This paper describes a method for automatic acquisition of wide-coverage treebank-based deep linguistic resources for Japanese, as part of a project on treebank- based induction of multilingual resources in the framework of Lexical-Functional Grammar (LFG). We automatically annotate LFG f-structure functional equations (i.e. labelled dependencies) to the Kyoto Text Corpus version 4.0 (KTC4) (Kurohashi and Nagao 1997) and the output of of Kurohashi-Nagao Parser (KNP) (Kurohashi and Nagao 1998), a dependency parser for Japanese. The original KTC4 and KNP provide unlabelled dependencies. Our method also includes zero pronoun identification. The performance of the f-structure annotation algorithm with zero-pronoun identification for KTC4 is evaluated against a manually-corrected Gold Standard of 500 sentences randomly chosen from KTC4 and results in a pred-only dependency f-score of 94.72%. The parsing experiments on KNP output yield a pred-only dependency f-score of 82.08%. Keywords: Lexical-Functional Grammar, Japanese, automatic linguistic resource acquisition, zero-pronoun identification 1. Introduction We present a method to automatically annotate Lexical-Functional Grammar (LFG)-style functional structure equations (labelled dependencies) on the unlabelled Kyoto University Text Corpus version 4 (KTC4) (Kurohashi and Nagao 1997), to acquire more abstract and (somewhat) less language-dependent LFG f-structure representations for Japanese sentences. We apply the algorithm to enrich the output of a Japanese dependency parser (Kurohashi- Nagao Parser, KNP) (Kurohashi and Nagao 1998), to construct f-structure representations for KNP output; the enriched parser output is available for further cross-linguistic research or applications such as machine translation. Our annotation method is based on the assumption that non-configurational, relatively free word-order languages, of which Japanese is one example, do not require phrase structure trees as an indispensable level of linguistic representation. Rather, the rich morphological information on each unit in a sentence, along with the unlabelled dependency between syntactic units in KTC4 and KNP output, provides us with as much information as what can be deduced from phrase-structure trees in other configurational, fixed word-order languages. Our method provides zero pronoun identification as a preliminary process for long distance dependency (LDD) resolution, based on the morphology of verbs and on the probability of subcategorization frames, associated with particular verbs. This paper has the following structure: in Section 2 we summarize the background of this research, including LFG and related work. In Section 3, we describe in detail our method of automatic annotation of f-structure functional equations on KTC4, and show how we approach the problem of zero-pronoun identification and present results of our f-structure annotation and * We gratefully acknowledge support from Science Foundation Ireland grant 04/IN/I527 for the research reported in this paper. Copyright 2007 by Masanori Oya and Josef van Genabith 375 parsing experiments. We discuss the overall results and their implications in Section 4, and conclude in Section 5. 2. Background 2.1 Lexical-Functional Grammar Lexical-Functional Grammar (LFG) (Bresnan 2001; Dalrymple 2001) is a syntactic theory in …

Content truncated to 500 words