UPAR7: A knowledge-based system for headline sentiment tagging

Proceedings of the 4th International Workshop on Semantic Evaluations (SemEval-2007), pages 422–425, Prague, June 2007. c⃝2007 Association for Computational Linguistics UPAR7: A knowledge-based system for headline sentiment tagging François-Régis Chaumartin Lattice/Talana – Université Paris 7 30, rue du château des rentiers - 75013 Paris - France fchaumartin@linguist.jussieu.fr / frc@proxem.com Abstract For the Affective Text task at SemEval- 2007, University Paris 7’s system first evaluates emotion and valence on all words of a news headline (using enriched versions of SentiWordNet and a subset of WordNet- Affect). We use a parser to find the head word, considering that it has a major im- portance. We also detect contrasts (be- tween positive and negative words) that shift valence. Our knowledge-based system achieves high accuracy on emotion and va- lence annotation. These results show that working with linguistic techniques and a broad-coverage lexicon is a viable ap- proach to sentiment analysis of headlines. 1 Introduction 1.1 Objectives The detection of emotional connotations in texts is a recent task in computational linguistics. Its economic stakes are promising; for example, a company could detect, by analyzing the blo- gosphere, people’s opinion on its products. The goal of the SemEval task is to annotate news headlines for emotions (using a predefined list: anger, disgust, fear, joy, sadness & surprise), and for valence (positive or negative). A specific difficulty here is related to the small number of words available for the analysis. 1.2 Overall architecture Our system is mainly rule-based and uses a lin- guistic approach. From a macroscopic point of view, we follow the hypothesis that, in a news title, all the words potentially carry emotions. If linguis- tic resources make it possible to detect these emo- tions individually, how can we deal with headlines where bad and good emotions appear at once? Our objective is to identify the expression which carries the main topic of the title. One can consider that this expression has a primary importance. We also seek to lay down rules for detecting specific emotions. For instance, surprise some- times comes from the contrast between good and bad news. And sometimes, simple lexical elements are characteristic of an emotion; a negation or a modal auxiliary in a title may be a relevant indica- tor of surprise. We describe here the techniques we imple- mented to address all these points. 2 Components & resources used The system we employed for the Affective Text evaluation consists of the following components1: • The SS-Tagger (a Part-of-Speech tagger)2, • The Stanford Parser. We also used several lexical resources: • WordNet version 2.1, • A subset of WordNet-Affect, • SentiWordNet. As the SS-Tagger is straightforward, we will not say more about it here. We will, however, discuss the remaining components and resources below. 1 We used them through the Antelope NLP framework (www.proxem.com), which makes them easy to use. 2 This fast PoS tagger uses an extension of Maximum Entropy Markov Models. See (Tsuruoka, Tsujii, 2005). 422 2.1 Choice of the Stanford Parser We wished to use a …

Content truncated to 500 words