Understanding users’ behavior mechanism of information diffusion on government microblogs is helpful for formulating effective strategies to promote public participation. However, limited effort has been made to examine the effects of extensive textual features and their differences on different topics. Therefore, on the basis of the elaboration likelihood model, we develop a model to explain how the extensive semantic features will influence the diffusion of government microblog posts with different topics. A model test with real data from Sina Weibo demonstrates the promotional effect of positive words, city names, adjectives/adverbs, and dissimilar contents, and that negative words will hinder the diffusion. The effect of city names for political news is greater than that for living information, while the influence of positive words for living information is greater than that for political news. Contributions to the literature and practice are discussed.

了解用户在政府微博上的信息传播行为机制，有助于制定有效的促进公众参与的策略。然而，研究大量文本特征的影响及其对不同主题的差异的努力有限。因此，在阐述似然模型的基础上，我们开发了一个模型来解释广泛的语义特征将如何影响不同主题的政府微博帖子的传播。新浪微博真实数据的模型测试表明，正面词、城市名称、形容词/副词和不同内容的宣传效果，负面词会阻碍传播。城市名称对政治新闻的影响大于对生活信息的影响，而正面话语对生活信息的影响大于对政治新闻的影响。讨论了对文献和实践的贡献。