Understanding interdependency among requirements is one of the success factors in software development. Information on requirements interdependency explicitly and implicitly resided various design artifacts or diagrams. A software requirements specification document is the artifact delivered in the early phase of development. It drives its following development processes. It also contains information on interdependencies among the requirements, such as similar, part-of, and elaborate. This study proposes an approach to model the requirement dependency graph for a software requirements specification document. There is an extraction process for Text Preprocessing, which includes of Tokenization, Stopword Removal, and Stemming. Besides, there is a process of measuring semantic similarity through WS4J (WordNet Similarity for Java). The results of the extraction process, combined with Greedy Algorithms as the optimum value solution approach. Besides, a method for calculating similarity was used through the practices of Wu Palmer and Levenshtein. At the end of this process, Reliability is performed using the Gwet's AC1 formula.