Towards Language Independent NE Identification in the context of Wikipedia

Named Entity Identification/Recognition is a key component for most Information Extraction tasks. All the existing approaches for NEI use extensive language specific resources. This paper deals with the problem of Multi lingual Named Entity Identification (NEI), and explains the need to address this problem in a language independent fashion. In this work we focus on Less Resourced languages like Hindi, Tamil that does not have prevalent language resources. Inherent structure of Wikipedia articles in multiple languages is exploited for NEI of less resourced languages. Other major contribution of this work is to extend the identification to word level from phrase level, under the intuition that it would increase the coverage of Named Entities. We evaluate our approach on comparable list of NE’s in Hindi, Tamil and English that are manually collected from Named Entity Workshop (NEWS).