Generation of Bilingual Dictionaries using Structural Properties

Authors

  • Ajay Dubey Search and Information Extraction Laboratory, International Institute of Information Technology.
  • Vasudeva Varma Search and Information Extraction Laboratory, International Institute of Information Technology.

DOI:

https://doi.org/10.13053/cys-17-2-1520

Keywords:

Bilingual dictionary, comparable corpora, structural elements.

Abstract

Building bilingual dictionaries from Wikipediahas been extensively studied in the area of computationlinguistics. These dictionaries play a crucial role inNatural Language Processing(NLP) applications likeCross-Lingual Information Retrieval, Machine Translationand Named Entity Recognition. To build thesedictionaries, most of the existing approaches useinformation present in Wikipedia titles, info-boxes andcategories. Interestingly, not many use the structuralproperties of a document like sections, subsections,etc. In this work we exploit the structural properties ofdocuments to build a bilingual English-Hindi dictionary.The main intuition behind this approach is thatdocuments in different languages discussing the sametopic are likely to have similar structural elements.Though we present our experiments only for Hindi, ourapproach is language independent and can be easilyextended to other languages. The major contribution ofour work is that the dictionary contains translation andtransliteration of words which include Named Entities to alarge extent. We evaluate our dictionary using manuallycomputed precision. We generated a massive list of 72ktokens using our approach with 0.75 precision.

Downloads

Published

2013-06-29