Reader Emotion Classification of News Headlines

Abstract

Emotion classification of text is very important in applications like emotional text-to-speech (TTS) synthesis, human computer interaction, etc. Past studies on emotion classification focus on the writer's emotional state conveyed through the text. This research addresses the reader's emotions provoked by the text. The classification of documents into reader emotion categories has novel applications. One of them is to integrate reader emotion classification into a web search engine to allow users to retrieve documents that contain relevant contents and at the same time produce proper emotions. Another is for websites to organize contents according to reader emotion categories and provide users a convenient browse. In this paper, we explore sentence level emotion classification. Firstly, we extract news headlines and related reader emotion information from the web. Then we classify news headlines into reader emotion categories using support vector machine (SVM), and examine classification performance under different feature settings. Experiments show that certain feature combinations achieve good results.