Automatically Characterizing Salience Using Readers' Feedback


  • Jean-Yves Delort Montpellier 2 University


Salience is an important characteristic of information influencing users’ cognitive and emotional states. For example, salient parts of a document are those that readers will find moving or provoking. This article studies the salience concept and its meanings in linguistics and information retrieval. Then it analyses the main drawbacks of content-based techniques for automatic identification of salient passages in a document. A new context-based method for overcoming these difficulties is subsequently presented. Our method identifies passages that readers have reacted to by analyzing their textual feedback. Our experimentation with blog posts revealed that it is effective and can be on 90% of commented posts.