Feedback requested on nlp project related to news story chains changing over time

Follow the full discussion on Reddit.
This sub has been very helpful in the past so I am hoping I can get some feedback. For my project, I am essentially trying to find a way to detect changes over time to a news narrative. At this stage I have applied an algorithm to successfully group together the stories that follow the development of the same event. So now I need to find a way to track, analyze, and maybe quantify how the events (and their coverage changes) . My current approach is using topic modeling to find important keywords in each of the articles. Then, I use those key words to map how the stories change overtime. So in the most basic of terms, I am identifying key words in the first article in each narrative chain and then comparing how those keywords change and are different from key words identified in subsequent articles about the same event. Does this approach sound reasonable? Is there anything else I should be trying instead? Thanks everyone!