Incorporating Structural Information in Scientific Document Retrieval

With the daily-increasing development of science, various methods have been designed to more and better retrieve the scientific documents based on the need and search of users. For some documents in the various scientific databases, no complete information exists and the users have to observe the inside of a document in order to catch up with its metadata inclusion the authors, their affiliations, the references cited and etc. Therefore, presence of a method based on extracting the information based on the available structural and geometrical properties in a document can assist the recovery of related and required documents. In addition, the available pitfall in the relational data based is the lack of direct and indirect relationships between the availabilities of each system for which a graph-oriented database can establish the relations between these availabilities. In this respect, after extracting metadata using the geometrical properties of document and using a graph-oriented model, the relations between various documents' availabilities such as authors, conferences, subjects and keywords and etc. are modeled in order to retrieve the information more effectively. The extracted data are refined and stored in the graph model and will be available for a user via a web-based user interface. To produce the results of each search, the related documents will be retrieved based on the graph relations and be weighed according to the rate of relatedness of each document and the number of references. In order to evaluate the proposed method, PubMed Database is used. The results of experiments show the proposed methods outperformed 60% in contrast to the PubMed Database search engine in terms of the retrieved documents. Furthermore, based in the F-measure, and nDCG-measure of proposed method considerably outperformed the PubMed Database search engine in terms of the quality of retrieved documents.