Semantic Processing of Compounds in Indian Languages

Proceedings of COLING 2012: Technical Papers, pages 1489–1502, COLING 2012, Mumbai, December 2012. Semantic Processing of Compounds in Indian Languages Amba Kulkarni1 Soma Paul2 Malhar Kulkarni3 Anil Kumar1 Nitesh Surtani2 1. Department of Sanskrit Studies, University of Hyderabad 2. International Institute of Information Technology, Hyderabad 3. School of Humanities and Social Sciences, I.I.T. Bombay apksh@uohyd.ernet.in, soma@iiit.ac.in, malharku@gmail.com, anil.lalit22@gmail.com, nitesh.surtani0606@gmail.com ABSTRACT Compounds occur very frequently in Indian Languages. There are no strict orthographic conventions for compounds in modern Indian Languages. In this paper, Sanskrit compounding system is examined thoroughly and the insight gained from the Sanskrit grammar is applied for the analysis of compounds in Hindi and Marathi. It is interesting to note that compounding in Hindi deviates from that in Sanskrit in two aspects. The data analysed for Hindi does not contain any instance of Bahuvr¯ıhi (exo-centric) compound. Second, Hindi data presents many cases where quite a lot of compounds require a verb as well as vibhakti(a case marker) for its paraphrasing. Compounds requiring a verb for paraphrasing are termed as madhyama-pada-lop¯ı in Sanskrit, and they are found to be rare in Sanskrit. KEYWORDS: Multi Word Expression, Compounds, P¯an.ini, As.t.¯adhy¯ay¯ı, semantics. 1489 1 Introduction Noun compounds represent a linguistic device of encrypting information that makes their analysis a challenging NLP task. For example, consider a compound ‘ballpoint pen’. It is made up of two nouns: a modifier ‘ballpoint’ and a head ‘pen’. The modifier is again a compound that is composed of two nouns ‘ball’ and ‘point’. The relation between ‘ball’ and ‘point’ and in turn between ‘ballpoint’ and ‘pen’ is not encoded anywhere on the surface although a user of the language can decodify the meaning correctly. The study of compounds involves two major tasks: 1) automatic identification and extraction of compounds from natural language texts and 2) syntactic and semantic analysis of compounds. The task of identification of compounds becomes significant because of orthographic vagaries. Orthographic conventions for writing compounds may vary from language to language and even within the same language. For example, in Sanskrit a compound is a single word, while in English (as exemplified above) as well as in modern Indian Languages (ILs) we find the following conventions of writing: components written with or without a space and components separated by a hyphen. When compounds are written without a space, the adjoining phonemes undergo euphonic changes as in ga˙ng¯a-udaka changing to ga˙ngodaka. Analysis of compounds primarily involves the expansion of these syntactically condensed constructs with an aim to unfold the meaning of the constructions (Butnariu and Veale, 2008; Girju et al., 2007; Kim and Baldwin, 2006; Kumar, 2012; Nakov, 2008; Nastase and Szpakowicz, 2009; Séaghdha and Copestake, 2007). Semantic analysis of compound is significant for various NLP applications including machine translation, information extraction and so on. Here we discuss one example to show how semantic analysis helps in machine translation. For example, let us consider the English compounds cancer death and room temperature. Lexical substitution of components of these compounds into Hindi would produce the …

Content truncated to 500 words