Sheffield-Trento System for Sentiment and Argument Structure Enhanced Comment-to-Article Linking in the Online News Domain

In this paper we describe and evaluate an approach to linking readers’ comments to online news articles. For each comment that is linked based on its comment, we also determine whether the commenter agrees, disagrees or stays neutral with respect to what is stated in the article, as well as what the commenter’s sentiment towards the article is. We use similarity features to link comments to relevant article segments and Support Vector Regression models for assigning argument structure and sentiment. Our results are compared to competing systems that took part in MultiLing OnForumS 2015 shared task, where we achieved best linking scores for English and second best for Italian.