UNIBA-CORE: Combining Strategies for Semantic Textual Similarity