This may be possible if you can find some kind of pattern that will allow you to identify the commas that are in the descriptions, and distinguish them from the others.
Can you do that ?
Or can you post maybe 2 or 3 examples of your data so that maybe people on this board can identify patterns ?
For example, if all the commas within descriptions were followed by a space character, and none of the others were, you could use Edit, Find and Replace, to find all the instances of ", " and replace them with some text string, let's say "zzz".
Then the remaining instances of "," would all be valid delimiters, and you could import the data.
Finally, you could use Find and Replace to find all the instances of "zzz", and replace them with ",".