Identifying the factors of country risk fluctuation from news text data using natural language processing