If your data is formatted that liberal, you are completely on your own. There ought to be rules for determining where fields/columns start and end. If there are no rules, you cannot parse. Period.

Is the current "format" the only possible format? Can the "data" be generated as something that does have rules, like CSV? When the data is well-formatted CSV, you can use Text::CSV_XS to parse the data and use all advice already given, or even easier, use Spreadsheet::Read (in combination with Text::CSV_XS) to get direct access to every "cell" in your dataset.