I think that's a little unfair. It really comes down to parsing text and you'll have similar issues even if you use a database or whatever you think the "real" solution is. I have a project I'm working on right now that stores dates, phone numbers, and website links. Cleaning/parsing has been 90% of the work and I still have edge cases that aren't fully solved. Every time I think I'm done, I find something else I haven't thought about. Local AI models have been a huge help though for sanitizing.