Web pages and their embedded documents are a good source of information. However, issuing complex queries directly on web pages using direct pattern-matching techniques are challenging tasks that require lengthy procedural programming. Also, programmers working directly on web pages are in charge of the search path and routing, which depend on individual document structures and affect the correctness, completeness, and performance of the results. In contrast, relational databases are well-structured and backed by mathematical principles. Well-designed relational database structures are known to be anomalies-free. The standard relational database language, SQL, is a non-procedural language that defines the required results precisely. Query results are both correct and complete. Performance issues are handled by intelligent query optimizers employed by modernday Database Management System (DBMS). This paper suggests an approach that transforms documents embedded on web pages in HTML format to corresponding relational database structures and populations. Functional dependencies (FDs) and multi-valued dependencies (MVDs) obtained from documents on the webpages are used to construct conceptual schema diagrams, which are further transformed into the Optimal Normal Form (ONF) relational database structures. In this research project, the Object Role Model (ORM) conceptual schema model is employed. The paper discusses the ORM and the rationales behind its usage. The detection of FDs and MVDs from webpage documents and the technical properties of the ONF relational database structures. Illustrated examples are also provided.