Multilingual lexical resources to detect cognates in non-aligned texts