Marathi/Hindi to English Transliteration of Named Entity using WEKA Tool

  Mayur Navsupe,  Manikrao Dhore

Phonetic property of a word is the property which describes the speech and pronunciation sound. Transliteration is a mapping of a word into one language from another language without losing its phonetic properties. Named entities (NE) are the words which are not found in the dictionary but they are used to support translation process. The process of transliteration of these named entity is always have been difficult. There are no predefined rules or regulation for transliteration which will not lead to loss of phonetic properties of named entities. People are writing different spellings at different places for the same name. This fact certainly affects the top-1 accuracy of the transliteration and in turn the transliteration process. We introduce the concept to use machine learning approach on NE data to solve this problem. We have used WEKA tool, a machine learning tool for training of data. We have used SMO classifier for the classification of NE data. We have applied set of linguistic rules on the predicted data for better results. Initial experiment shows use of machine learning tool and SMO is very helpful for transliteration process.

Keywords- Transliteration, Named Entity, WEKA, SMO (Sequential minimal optimization), Linguistic.

Unique Identification Number - IJEDR1602256
Page Number(s) - 1436-1442
Pubished in - Volume 4 | Issue 2 | June 2016
DOI (Digital Object Identifier) -   
Publisher - IJEDR (ISSN - 2321-9939)

  Mayur Navsupe,  Manikrao Dhore,   "Marathi/Hindi to English Transliteration of Named Entity using WEKA Tool", International Journal of Engineering Development and Research (IJEDR), ISSN:2321-9939, Volume.4, Issue 2, pp.1436-1442, June 2016, Available at :http://www.ijedr.org/papers/IJEDR1602256.pdf