Semantic Net Based Hindi Information Retrivel System*

(आर्थीय-तंत्र आधारित हिन्दी सूचना प्रत्यानयन-प्रणाली)

-अरिमर्दन कुमार त्रिपाठी

Information Retrieval is one of the most challenging and enormous fields of Natural Language Processing. In the present social scenario 'information' is treated as not only knowledge but also as constituents of power in its context with its own property.In this corpus, which represents natural language into language oriented computational behaviors; syntax is the key feature for information point of view, which can be covered through some punctuation marks, available in the texts. For retrieving information from any text, the supportive database of a hierarchically mapped semantic net should be prepared in concerned languages, and the corpus should be POS and syntactically tagged.The role of a semantic net is central in this process, therefore it's database will be normalized. The result query will cover all the probable results despite mistakes committed by humans, like spelling and orthographic variations (etc.). In real time processing, the first query will go into the semantic net, then it will match the relevant text according to that query. That is why a hierarchically mapped and normalized database will be more helpful to get results with sorted frame.Although in this model preprocessing is required within the certain adopted framework at several levels, such as tagging, normalizing and hierarchical mapping of data as per its use and role in society. In its result, however, there is no need of post processing; the result will be indexed according to the mapped hierarchy so that output will be relevant and well matched to queries.In this model, stemming and ambiguity resolution of a given query will be helpful in matching. Codification of stop words will increase the speed of system processing.
*Abstract of M.Phill. Research work