WACK: Advancing Hallucination Detection by Identifying Knowledge-Based Errors in Language Models Through Model-Specific, High-Precision Datasets and Prompting Techniques