Mapping and Structural Analysis of Multi-lingual Wordnets

by J. Ramanand. Akshay Ukey. Brahm Kiran Singh. Pushpak Bhattacharyya.

In this paper, we present observations on structural properties of wordnets of three languages: English, Hindi, and Marathi. Hindi and Marathi, spoken widely in India, rank 5th and 14th respectively in the world in terms of the number of people speaking these languages. The observations suggest the existence of the ‘small world’ property in wordnets and also lend credence to the belief that the world of concepts, which the words are manifestations of, share common properties across languages. These concepts are represented by synsets (sets of synonymous words) in wordnets. Therefore, it makes sense to link the synsets of wordnets of different languages to create a global wordnet grid. In fact, the EuroWordnet project is already doing so for a number of European languages. We too report our work on linking English, Hindi and Marathi synsets. The first task – linking of English and Hindi wordnets – requires clever ideas on mapping accurately the synsets of the two languages. The second task – relatively easier due to the close correspondence between Hindi and Marathi – reduces to programmatically borrowing lexico-semantic relations from the Hindi wordnet into the Marathi wordnet, since the Marathi wordnet is largely aligned with the Hindi wordnet. To the best of our knowledge, ours is the first in-depth investigation into lexical knowledge networks of multiple languages with a view to evaluating them, and also the first attempt to create a multiwordnet involving two major Indian languages and English. [via]

ftp://ftp.research.microsoft.com/pub/debull/A...

Download File

PDF Finder

Mapping and Structural Analysis of Multi-lingual Wordnets

Related Files

Rate this Document

Popular eBooks

ADS

Tag Clouds

Last Download

BookShelf