Semantically Motivated Hebrew Verb-Noun Multi-Word Expressions Identification

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 1242–1253, Osaka, Japan, December 11-17 2016. Semantically Motivated Hebrew Verb-Noun Multi-Word Expressions Identification Chaya Liebeskind, Yaakov HaCohen-Kerner Department of Computer Science, Jerusalem College of Technology, Lev Academic Center 21 Havaad Haleumi St., P.O.B. 16031 9116001 Jerusalem, Israel liebchaya@gmail.com, kerner@jct.ac.il Abstract Identification of Multi-Word Expressions (MWEs) lies at the heart of many natural language processing applications. In this research, we deal with a particular type of Hebrew MWEs, Verb- Noun MWEs (VN-MWEs), which combine a verb and a noun with or without other words. Most prior work on MWEs classification focused on linguistic and statistical information. In this paper, we claim that it is essential to utilize semantic information. To this end, we propose a semantically motivated indicator for classifying VN-MWE and define features that are related to various semantic spaces and combine them as features in a supervised classification framework. We empirically demonstrate that our semantic feature set yields better performance than the common linguistic and statistical feature sets and that combining semantic features contributes to the VN-MWEs identification task. 1 introduction Multi-word expressions (MWE) were defined by Sag et al. (2002) as “idiosyncratic interpretations that cross word boundaries (or spaces)” while Bouamor et al. (2012) defined a MWE as “a combination of words for which syntactic or semantic properties of the whole expression can not be obtained from its parts”. Jackendoff (1997) claimed that that the frequency of MWEs in a speaker’s lexicon is of the same order of single words. Due to their relative high frequency and complexity, MWEs require high-quality treatment in many applications in natural language processing (NLP) such as data mining, machine translation (MT), information retrieval, natural language understanding, natural language generation, question answering (QA), text summarization, and word sense disambiguation (WSD). The aim of this work is to explore Hebrew Verb-Noun MWEs (VN-MWEs). VN-MWEs are MWEs whose constitutents include a verb and a noun. The motivation of this research is to enable automatic identification of VN-MWEs for various NLP tasks such as MT, QA, and WSD, and to classify colloca- tions that include verbs as VN-MWEs or non-VN-MWEs. Most prior efforts to automatically classify MWEs focused on three approaches: (1) Statistical ap- proaches, either frequency-based or co-occurrence-based (Dias et al., 1999; Deane, 2005; Pecina and Schlesinger, 2006). (2) Linguistic approaches that are based on NLP tools, such as taggers and parsers (Al-Haj, 2009; Bejcek et al., 2013; Green et al., 2013). (3) Hybrid approaches which combine statistical and linguistic approaches (Baldwin, 2005; Boulaknadel et al., 2008; Farahmand and Nivre, 2015). In this paper, we claim that on top of standard linguistic and statistical metrics, MWE identification methods can greatly benefit from exploiting semantically motivated cues. For example, when a VN- MWE is highly idiomatic, the semantics of the verb and the noun are not likely to overlap. The contribution of this paper is, in a first step, to combine semantic features in the framework of supervised MWE classification. We suggest …

Content truncated to 500 words