Show, Don't Tell: Uncovering Implicit Character Portrayal using LLMs