Detecting Hallucinations in Large Language Model Generation: A Token Probability Approach