You record all upcoming changes, store them with a useful key (in this case the name of the book or chapter) and when stopping the record you'll have a set of new information referring to the book you read. Then you do it again with the next book or chapter and so on.
But not only books! It could be any kind of source or relevant hint for the tag of the recorded set (e.g. Date, Lecture, Grade etc.). One node should have 1:n relevation to the records because the information / node could be given in another book too.