Proceedings of the Annual Meeting of the Cognitive Science Society

The challenge of predict ing meme success has gained at tention from researchers, largely due to the increased availability of social media data. Many models focus on st ructural features of online social networks as predictors of meme success. The current work takes a different approach, predict ing meme success from linguist ic features. We propose predict ive power is gained by grounding memes in theories of working memory, emot ion, memory, and psycholinguist ics. The linguist ic content of several memes were analyzed with linguist ic analysis tools. These features were then t rained with a mult ilayer supervised backpropagat ion network. A set of new memes was used to test the generalizat ion of the network. Results indicated the network was able to generalize the linguist ic features in order to predict success at greater than chance levels (80% accuracy). Linguist ic features appear to be enough to predict meme t ransmission success without any informat ion about social network st ructure.