My weekend task: extracting sensible machine-readable text from a scientific journal article. Of course, articles are published using PDF - so any text flow has been completely garbled. Why do we do this in 2022?