Named Entity Recognition in Urdu: A Progress Report

摘要

We are interested in properly identifying named entities in Hindi and Urdu text for natural language processing purposes, including information extraction. We explore two approaches to processing Hindi and Urdu text and designing named entity recognition algorithms. In the first approach, using the Unicode character set, we consider processing Hindi in Devanagari script and Urdu in Arabic script instead of transcribing the languages to a Roman-based script. In the second approach, we consider transcribing the Hindi and Urdu text to a common script, the International Phonetic Alphabet (IPA). As part of the project, we built an Urdu corpus marked up with an Extensible Markup Language (XML). We consider both statistical-based and rule-based approaches to the named entity recognition algorithms.