Personalized text snippet extraction using statistical language models