Predicting News Headline Popularity with Syntactic and Semantic Knowledge Using Multi-Task Learning

AbstractNewspapers need to attract readers with headlines, anticipating their readers’ preferences. These preferences rely on topical, structural, and lexical factors. We model each of these factors in a multi-task GRU network to predict headline popularity. We find that pre-trained word embeddings provide significant improvements over untrained embeddings, as do the combination of two auxiliary tasks, news-section prediction and part-of-speech tagging. However, we also find that performance is very similar to that of a simple Logistic Regression model over character n-grams. Feature analysis reveals structural patterns of headline popularity, including the use of forward-looking deictic expressions and second person pronouns.