Rule-Based Recognition of Associated Entities in Hindi Text: A Domain Centric Approach

There are several computing applications that process unstructured texts available in various natural languages on the Web. These applications provide worthwhile outcomes that are of interest to the end users, other application developers and researchers. Various approaches of Natural Language Processing (NLP) and related fields are used to develop these applications. Application’s domain related lexical resources, in same language as the language of the text being processed, are considered useful and effective in improving performance of the related application. Several such resources are available in English, but there are certain languages that have scope for development and research. There is a need to develop the resources in those languages and bring them at par with their English counterparts. One such identified language is Hindi. Lexical resources in Hindi are developed in an ongoing research in the field of Opinion Mining. These are used in a rule based named entity recognition system developed as part of the research. The unstructured texts from Hindi Weblogs in the domain of home remedies are used in this research. The methodology used to develop these resources, outcomes achieved, experimental setup used for evaluation of the developed resources and the results obtained thereof, are discussed in this paper.