Agfdhyk

I want to analyse tables with similar data, that are structured differently and where the headers also may be slightly diverse.

For collecting all the data from the tables summing them up I face several problems.

Step 1: I look for the header keywords. Searching for if "cars==cars" is not possible, because the header may appear as "car", "Car" or "Cars". There is also the possibilty that there is a spelling mistake in the word. So iterating through all possibilites can also result in false.
When I search for solutions to this problem I found out about the fuzzy logic, but I would be thankful about other approaches.

Step 2: I found the desired keyword in the table, but how do I know where the related data is placed? It can be below it, but also right cell next to it. Are there approaches to get information about the general structure of the table?

Step a (part 1) - naive implementation would be dictionary distance (as you want to handle typos)

Step a (part 2) - use synonym database / thesaurus to find similarly named columns

Step b (part 1) - data is aligned the same way the headers are - so if headers are aligned vertically, then data will be as well

Step b (part 2) - similar data will have the similar data type (raw string, number, zip-code), by checking to the right and to downwards you can detect which is the real direction.