Learning Recognition of Ambiguous Proper Names in Hindi

Abstract

An ambiguous proper name is a name which is also a valid dictionary word with a meaning of its own when used in the text. For example in English, the word 'bush' in 'Mr. Bush' is a proper name whereas in 'a dense bush' it is a lexical entity. Almost all proper names in Hindi have a meaning and find an entry in the dictionary. Recognition of named entities finds wide application in MT, IR and several other NLP tasks. While there have been a number of investigations on Hindi NER in general, no work has been reported exclusively on ambiguous proper nouns which are more difficult to deal with. This paper presents a methodology for recognizing ambiguous proper names in Hindi using hybridization of a rule-base and statistical CRF based machine learning using morphological and context features. The methodology yields a F-score of 71.6%.