The problem is to search the 'pdesc' column of 'part_of_car' table containing
the text of each of value of the 'descriptiontext' of the 'category' table and UPDATE 'part_of_car'.'pcategory' column with the corresponding 'keyword' column of the 'category' table.
The UPDATED 'part_of_car' Table is expected to be :
What if part_of_car.pdesc contains two (or more) category.descriptiontexts? For example, "1764-FJ-3 ENGINE BRAKE PAD KIT" (never mind this example means nothing in real life). What would you take as a result for part_of_car.pcategory? ENG or BRK?
If those tables contain a few data, I guess result will be more accurate if you do it yourself (scroll through data and mark them).
But, if it is impossible, I'm afraid you might have a table full of garbage.
However ... I'd say you need two nested loops: first would read part_of_car table and extract, as substrings, words of pdesc column ("110-702-3S WATER PUMP " = "110-702-3S" + "WATER" + "PUMP") and then, in another loop, compare that substring with all the category.descriptiontext values. If they match, update part_of_car table with the category.keyword you've found.
Quite simple, actually ... If that's what you need, of course