Evolving conversions between trees (representing JSON) and graphs (representing relational database schemas)

I have in input different, say 100, types of trees with labeled nodes (representing JSON files). I need to transform the information contained in the trees into graphs with labeled nodes (representing insert statements into tables in a relational database).

The structure of the trees evolves very quickly. Every month some labels are moved, renamed, or their type changes.

Let’s say that I know how to transform trees into graphs at time t, is is possible to infer somehow at least part of the new transformations with the new tree structure?

Are there papers or books that I need to read to know some theory that could help me to tackle this task?