Enhancement of Query Refinement for Marathi Language Information Retrieval

Introduction: The Information Retrieval (IR) enables users to access relevant data in a language different from their queries. The IR involves the capability to submit a query in one language and retrieve documents in the same or another language, like Marathi or English. This IR is accomplished by creating a system that compares a query in one language with data in the same or another language. Objectives: This study presents a novel approach to enhance query refinement for Marathi language information retrieval, focusing on addressing the language's unique linguistic complexities. Methods: proposed SVM model of Marathi language query refinement model over the Marathi Dataset. The dataset is preprocessed using tokenizing, pos-tagged, and lemmatized information. after the processing, The Rocchio algorithm retrieves the relevant query results from the vector space model. The fine-tuned SVM trains to distinguish between relevant and irrelevant documents. Results: The SVM achieved an accuracy score of 0.79, shows to correctly classify documents as relevant or irrelevant. While the classifier demonstrated strong performance in identifying irrelevant documents, there is room for improvement in correctly identifying relevant documents. Conclusions: The proposed model shows potential for improving the accuracy score and efficiency of information retrieval in the Marathi language, contributing to the advancement of NLP techniques for less-resourced languages.