This paper proposes a method for generating news headlines that include a specific person’s name from the news article. We adopt T5, a modern large language model (LLM), for news headline generation. Conventional news headline generation methods input the entire article into the LLM. Attempting to summarize the entire article in the limited number of characters of a news headline may generate a sentence with a meaning that is not factually correct. Therefore, we propose a method that only inputs meaningful sentences extracted from the news article. The proposed method extracts two types of sentences: a sentence represents the article’s main topic, and a sentence contains facts about the given specific person’s name. Through evaluation experiments, our method demonstrated better accuracy in generating news headlines focusing on a specific person than methods inputting the entire article.