Automation in the legal domain is promising to be vital to help solve the backlog that currently affects the Indian judiciary. For any system that is developed to aid such a task, it is imperative that it is informed by choices that legal professionals often take in the real world in order to achieve the same task while also ensuring that biases are eliminated. The task of legal case similarity is accomplished in this paper by extracting the thematic similarity of the documents based on their rhetorical roles. The similarity scores between the documents are calculated, keeping in mind the different amount of influence each of these rhetorical roles have in real life practices over determining the similarity between two documents. Knowledge graphs are used to capture this information in order to facilitate the use of this method for applications like information retrieval and recommendation systems.