Creating a contemporary corpus of similes in Serbian by using natural language processing