Construction of an English Dependency Corpus incorporating Compound Function Words

Construction of an English Dependency Corpus incorporating Compound Function Words Akihiko Kato, Hiroyuki Shindo, Yuji Matsumoto Nara Institute Science of Technology 8916-5 Takayama, Ikoma, Nara 630-0192, Japan kato.akihiko.ju6@is.naist.jp, shindo@is.naist.jp, matsu@is.naist.jp Abstract The recognition of multiword expressions (MWEs) in a sentence is important for such linguistic analyses as syntactic and semantic parsing, because it is known that combining an MWE into a single token improves accuracy for various NLP tasks, such as dependency parsing and constituency parsing. However, MWEs are not annotated in Penn Treebank. Furthermore, when converting word-based dependency to MWE-aware dependency directly, one could combine nodes in an MWE into a single node. Nevertheless, this method often leads to the following problem: A node derived from an MWE could have multiple heads and the whole dependency structure including MWE might be cyclic. Therefore we converted a phrase structure to a dependency structure after establishing an MWE as a single subtree. This approach can avoid an occurrence of multiple heads and/or cycles. In this way, we constructed an English dependency corpus taking into account compound function words, which are one type of MWEs that serve as functional expressions. In addition, we report experimental results of dependency parsing using a constructed corpus. Keywords: MultiWord Expressions, Dependency Parsing 1. Introduction Multiword expressions (MWEs) consist of groups of to- kens, which should be treated as a single syntactic or se- mantic unit. MWEs are also known as “idiosyncratic inter- pretations that cross word boundaries” (Sag et al., 2002). The recognition of MWEs in a sentence is important for such linguistic analyses as syntactic and semantic parsing, because it is known that combining an MWE into a sin- gle token improves accuracy for various NLP tasks, such as dependency parsing (Nivre and Nilsson, 2004), and con- stituency parsing (Arun and Keller, 2005). For syntactic parsing that takes in MWEs, it is preferable that the information of an MWE (e.g., part of speech and span of tokens) be integrated into a corpus, such as a phrase or dependency treebank, because an MWE should be a syn- tactic unit. Actually, an MWE is grouped under “subtree” in French Treebank (Abeill´e et al., 2003), which is often used in research focusing on both MWE recognition and syntactic parsing (Green et al., 2011; Candito and Con- stant, 2014). However, MWEs are not annotated in Penn Treebank, the standard corpus of English syntactic parsing. In dependency structure that takes MWEs into consid- eration (i.e., MWE-aware dependency), each MWE be- comes a single node. Therefore, to convert word-based de- pendency to MWE-aware dependency directly, one could combine nodes in an MWE into a single node. Never- theless, this method often leads to the following problem: A node derived from an MWE could have multiple heads and the whole dependency structure including MWE might be cyclic (Figure 1). This is mainly because Penn Tree- bank style annotation does not give the special treatment for MWEs. We discuss this problem further in Section 3. In order to tackle the above problem, we …

Content truncated to 500 words