Using Feedforward and Recurrent Neural Networks to Predict a Blogger’s Age

Predicting the age of a blogger based on the text of their writing is a difficult task due to the fluidity of age identity and the effect of aging on writing styles. We propose feedforward and recurrent neural network frameworks to address this problem without enforcing human-generated features and find that shallow networks suffice for this problem. Results suggest that a scaled bag-of-words feedforward neural network model is better suited for age prediction than a pre-processed stacked long short-term memory model, possibly because word choice may be more indicative of author age than the meaning of the text.