Enriching Statistical Translation Models Using a Domain-Independent Multilingual Lexical Knowledge Base