Re^7: Module for parsing tables from plain text document

I stand corrected.

As for those advanced heuristics, my first instinct would be to look into the "Open/Import" functionality of all those open source Spreadsheet tools like LibreOffice. Those developers spent the last few decades writing software that can make sense of user provided, badly formatted data files.

As far as it concerns myself, those self-"learning" AI/heuristics/statistics tools might be somewhat interesting for occasional hobby use. But i wouldn't consider them for production use. If something goes wrong (e.g. "a bug happens"), it's easy enough to debug (and verify/certify) a handcrafted parser. If an AI goes wrong, all you can do is tweak the training data, retrain the model and pray to a $DEITY of your choice that

this has fixed the current problem
the change in your training data hasn't introduced new problems

Advanced statistics (including what we commonly refer to AI) is an amazing tool by itself. But when is goes wrong, you basically have to find an error (or omission) in what boils down to a formula with possibly tens of millions of variables. I mean, winning a Nobel price is nothing to sneer at, but i'm not sure how one would do it on a typical IT department budget ;-)

PerlMonks XP is useless? Not anymore: XPD - Do more with your PerlMonks XP

Comment on Re^7: Module for parsing tables from plain text document

Replies are listed 'Best First'.
Re^8: Module for parsing tables from plain text document by LanX (Saint) on Jan 13, 2023 at 15:14 UTC
> my first instinct would be to look into the "Open/Import" functionality of all those open source Spreadsheet tools like LibreOffice. I already mentioned the Excel's import wizard, but it's mainly meant for CSV and not free-form tables. > If an AI goes wrong, all you can do is tweak the training data, retrain the model and pray to a $DEITY of your choice that As I said, that's also not the way I would go. The AI should a TK window with a preview with the interpretation(s) and options to improve create a "wizarded" Perl code This code should be validating too, hence be more fault tolerant the most hand crafted code. (like detecting unusual data, like text in a field which always used to be numerical) Ideally the Perl code could also include data to restart the Tk-Wizard to refince in case of errors. I think that would easily address all your concerns. PS: Of course this could also be a realized as a web service and skip the Tk part. Cheers Rolf _{(addicted to the 𐍀𐌴𐍂𐌻 Programming Language :) Wikisyntax for the Monastery}	[reply]

Replies are listed 'Best First'.

Re^8: Module for parsing tables from plain text document
by LanX (Saint) on Jan 13, 2023 at 15:14 UTC

> my first instinct would be to look into the "Open/Import" functionality of all those open source Spreadsheet tools like LibreOffice.

I already mentioned the Excel's import wizard, but it's mainly meant for CSV and not free-form tables.

> If an AI goes wrong, all you can do is tweak the training data, retrain the model and pray to a $DEITY of your choice that

As I said, that's also not the way I would go.

The AI should

a TK window with a preview with the interpretation(s) and options to improve
create a "wizarded" Perl code

(like detecting unusual data, like text in a field which always used to be numerical)

Ideally the Perl code could also include data to restart the Tk-Wizard to refince in case of errors.

I think that would easily address all your concerns.

PS: Of course this could also be a realized as a web service and skip the Tk part.

Cheers Rolf
_{(addicted to the 𐍀𐌴𐍂𐌻 Programming Language :)

Wikisyntax for the Monastery}

[reply]