Keyword table and load mechanism should be updated to allow "duplicate" Japanese keywords to be stored

In testing the data loader it was noticed that some Japanese keywords are currently treated as duplicates in the db (and load mechanism). Nate said:

Ideally we would be able to keep them.

The toggle setting “include close variants for exact match” or whatever it’s called is set to “do not include” by default for AML, correct? If so, then it’s more important to keep these synonyms. Sachi’s research indicates that these fall under the category of things that would get mapped via exact if that toggle is set to “include”, so since we determined (I think) to set the toggle the other way, we’d need these kinds of things if possible. We’d still capture the traffic via broad-match keywords, but it’s better for overall performance to allow them to be separate keywords.

Thanks, Nate

The keywords database table probably needs to use a different collation and the load mechanism's temp table probably would as well.