One of the advantages of the language model approach to CLIR tasks is to enable us to put translation
probability p(tjs) directly into the formula where s is a source term and t is a target term. There are two
ways to incorporate the translation probability: