Influence and information diffusion is an important field in social network analysis. Diffusion models have a key role for predicting propagation behaviour of messages and are a cruicial prerequistic in solving other social network problems . In this research, we focus on propagation models applicable to the Influence Maximization Problem, particularly the widely used Threshold and Cascade models along with their various extensions. These models attempt to simulate how users adopt information based on network structure and peer influence.

We evaluate the effectiveness of these models in predicting the propagation behavior of tweets within the Twitter network. Our analysis reveals that while each model can accurately predict the spread of certain tweets, they fail to generalize across diverse content types. This inconsistency suggests that message content significantly influences propagation dynamics. To address this limitation, we propose two novel content-aware generalized propagation models that incorporate message semantics into the prediction process. These models dynamically select the most suitable propagation mechanism based on the tweet’s content characteristics. Experimental validation using real-world Twitter datasets demonstrates that our approach significantly improves prediction accuracy and offers a more robust framework for modeling information diffusion. Our findings highlight the importance of integrating content analysis into diffusion modeling for more effective influence strategies.