I am writing as part of my master thesis work an analysis of developers' work. I am looking for a new parameters to be consumed in Topological Data Analysis process. I would like to compare a similarities of changes made through two different commits. I can gather a commit message of these commits and differences in the source code. However, I don't know how good to count such similarity. I am considering to put message and text differences from source code into DBPedia Spotlight and count cosine similarity from the result annotations. Would you have some better idea of how to proceed it?