Internal matching algorithms are English-language oriented with unexamined heuristics

Where?

What's the situation?

What does it make more difficult?

How can we improve it?