A Text Formalization based Approach for Improved Query Response

Abstract

Information retrieval (IR) means automatic retrieval of required information into a wellorganized form from different sources. Early retrieval methods were ineffectual to beat the scalability and adaptability of natural language text, but with the arrival of internet, the scope and variance of applications that depends on diverse information retrieval and extraction has raised noticeably. As our society is becoming more information dependent, people need a way to easily access the information available for their needs. Applications such as inventory systems, medical decision support systems and institutional databases that keep the records of their day to day activities lead the practitioners to further research. To meet with this need, many approaches have been put into operation to catch the important information from text and make them accessible for decision making but the results are still substandard. In this thesis a new model for rules retrieval is presented. This model is based on Latent Dirichlet Allocation (LDA). The proposed LDA model is applied to a corpus of XML documents yielding topic terms. Using these topic terms as indexing terms brings significant improvement in the results as compared to alternative models.