According to the latest surveys by the Reuters Institute for the Study of Journalism, social media has become the primary source of news for most. Traditional news providers, such as newspapers and TV channels, have been leveraging social media to communicate with their audiences. Public media often provides more neutral news content compared to private-owned media. However, public media are usually less popular for the same reason, as careful word choice and the emphasis on a neutral perspective do not translate well to content popularity. In this study, we try to approach the problem of public media’s social media content popularity using text-mining techniques. We collected social media posts from Taiwanese public media and analyzed their content. Performance indicators for individual social media posts were also collected, including likes, comments, and shares. The two findings of the current study are as follows. First, The engagement from social media users seems to be related to the topic of content. The social media editor can strategically select appropriate topics for posting to induce different kinds of engagements. Second, the current results support that the three kinds of engagements are different not just in the magnitude of involvement, but also in the ways to motivate the audience to engage. For the news agencies that still follow traditional news writing style, it is still possible to improve the audiences’ social media engagement by choosing suitable topics.