Hack of the day: When working with ordered datasets, the GNU diff tool can be used for all kinds of filtering and formatting by different combinations of --unchanged-line-format, --old-line-format and --new-line-format. "%L" includes the line in question, and you can omit any output by specifying the given format as "".
For example, only gives lines that the new file omits, but prefixes them with "deprecated_":
diff --old-line-format="deprecated_%L" --unchanged-line-format="" --new-line-format="" old.txt new.txt
There are dedicated tools that allow doing such dataset comparisons, of course, but sometimes you might end up with a limited environment with just very basic tools.