We tried this a few years ago, but it is greatly problematic due to how language works.
Romance languages have lots of overlap, and even words cherry picked for a filter will have usages.
We currently do this for Han (Chinese) Unicode ranges, but that's one of the only cases where it's this clear-cut. Even then, it breaks Japanese, but that will be fixed next update.