Indian Language Wordnets and their Linkages with Princeton WordNet

Indian Language Wordnets and their Linkages with Princeton WordNet Diptesh Kanojia1,2,3, Kevin Patel1, Pushpak Bhattacharyya1 1IIT Bombay, 2Monash University, 3IITB-Monash Research Academy, {diptesh, kevin.patel, pb}@cse.iitb.ac.in Abstract Wordnets are rich lexico-semantic resources. Linked wordnets are extensions of wordnets, which link similar concepts in wordnets of different languages. Such resources are extremely useful in many Natural Language Processing (NLP) applications, primarily those based on knowledge-based approaches. In such approaches, these resources are considered as gold standard/oracle. Thus, it is crucial that these resources hold correct information. Thereby, they are created by human experts. However, human experts in multiple languages are hard to come by. Thus, the community would benefit from sharing of such manually created resources. In this paper, we release mappings of 18 Indian language wordnets linked with Princeton WordNet. We believe that availability of such resources will have a direct impact on the progress in NLP for these languages. 1. Introduction Wordnets (Fellbaum, 1998) have been useful in different Natural Language Processing applications such as Word Sense Disambiguation (TufiS¸ et al., 2004; Sinha et al., 2006), Machine Translation (Knight and Luk, 1994) etc. Linked Wordnets are extensions of wordnets. In addition to language-specific information captured in constituent wordnets, linked wordnets have a notion of an interlingual index, which connects similar concepts in different lan- guages. Such linked wordnets have found their application in machine translation (Hovy, 1998), cross-lingual informa- tion retrieval (Gonzalo et al., 1998), etc. Given the extensive application of wordnets in different NLP applications, creation and maintenance of wordnets involve expert involvement. Such involvement is costly both in terms of time and resources. This is further am- plified in case of linked wordnets, where experts need to have knowledge of multiple languages. India is a vast country with massive language diversity. Ac- cording to a census in 2001, there are 122 major languages 1, out of which, 29 have more than a million native speak- ers. The IndoWordNet project contains wordnets of 18 of these languages. These wordnets were created using expan- sion approach with Hindi Wordnet as the pivot. This paper makes the following contributions: • We release the latest version of 18 wordnets under the IndoWordNet project as a single bundle2. • Using mappings between Princeton WordNet and Hindi wordnet, we create and release mappings be- tween Princeton WordNet and these 18 languages wordnet. The rest of the paper is organized as follows: Section 2. covers some background and related work needed for fur- ther discussions. Section 3. describes the released re- sources. Section 4. discusses different issues encountered in the creation of these datasets, followed by the conclusion and future work. 1http://en.wikipedia.org/wiki/Languages of India 2http://www.cfilt.iitb.ac.in/ilw 2. Background and Related Work Princeton WordNet or the English WordNet was the first wordnet and inspired the development of many other word- nets. EuroWordNet (Vossen and others, 1997) is a linked wordnet comprising of wordnets for European languages, viz, Dutch, Italian, Spanish, German, French, Czech and Estonian. Each of these wordnets is structured in the same way as the …

Content truncated to 500 words