Another thought is that you could use an APPEND query based on a raw import of the spreadsheet to an INTERMEDIATE table.
First, do the raw import to a new table.
Then run DELETE queries on records that violate any of your screening laws. Like for example, DELETE * FROM TEMPTABLE WHERE [MainWord]=[Homonym] ;
You can also do UPDATES on the temp table to upcase or downcase everything in the mainword and homonym columns. And Trim$() for the words, too.
In other words, massage the holy Hell out of the TEMP table. When you are done, then and only then, import it to the final destination.
If you are going to index the destination table, my suggestion stops you from thrashing the crud out of the indexes so badly that you need to compact the result before you even try to use it. Indexes aren't needed in the TEMP table if you are doing (1) row-local updates and testing, and (2) simple edits for cases and spaces. Doing the append to the table with indexes already defined ONLY after you've done the filtration will make the result easier and cleaner to load. And less likely to give you key violation errors.
You'll still have to be careful, though, because many homonyms are more than duals. Like, maybe "their" and "there" and "they're" ?