Political News Sentiment Analysis for Under-resourced Languages

Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 2989–2996, Osaka, Japan, December 11-17 2016. Political News Sentiment Analysis for Under-resourced Languages Patrik F. Bakken, Terje A. Bratlie, Cristina Marco, Jon Atle Gulla Norwegian University of Science and Technology Trondheim, Norway bakken.patrik@gmail.com, terje.a.bratlie@gmail.com, {cristina.marco,jag}@ntnu.no Abstract This paper presents classification results for the analysis of sentiment in political news articles. The domain of political news is particularly challenging, as journalists are presumably objective, whilst at the same time opinions can be subtly expressed. To deal with this challenge, in this work we conduct a two-step classification model, distinguishing first subjective and second positive and negative sentiment texts. More specifically, we propose a shallow machine learning approach where only minimal features are needed to train the classifier, including sentiment-bearing Co- Occurring Terms (COTs) and negation words. This approach yields close to state-of-the-art results. Contrary to results in other domains, the use of negations as features does not have a positive impact in the evaluation results. This method is particularly suited for languages that suffer from a lack of resources, such as sentiment lexicons or parsers, and for those systems that need to function in real-time. 1 Introduction In the rapidly changing World Wide Web, getting informed opinions about facts is becoming increasingly challenging for users. In online news, the same event can be presented from very different perspectives depending on the source. Journalism is supposed to be objective, yet opinions are also expressed in newswire text (Belyaeva and van Der Goot, 2008; Blaz et al., 2009). In this type of texts, opinions are generally expressed in a subtle way, by using language resources other than sentiment vocabulary, such as irony, sarcasm, metaphors, etc., as it is illustrated in the following example, where the use of an ironic comparison som rundingsbøyer ‘as if they were human buoys’ suggests a negative opinion: - Den nye forskriften betyr at i vannene der friluftsfolket fortsatt kunne f˚a være i fred, i en liten bortgjemt vik, der er det n˚a ˚apent for skuterfolket ˚a bruke turfolket som rundingsbøyer, sier Bjørn Hansen, p˚a vegne av Naturvernforbundet i Finnmark. (- Given by the new regulation, places near the waters where people still could have piece and quiet, such as small secluded coves, are now open for people on jet skis to make use of the former as if they were human buoys, says Bjrn Hansen, on behalf of the Nature Conservatory of Finnmark.)1 In this context it is difficult for readers to have informed opinions about the events happening in the world. Thus there is a growing need for resources that can help readers filter out news information, so that they can have an informed opinion about the current events. These resources would be particularly useful in the domain of political news, where readers want to be informed about political parties, politicians, and policies. If successfully employing sentiment analysis in the political news domain, possible biases and opinions will be revealed, and readers will …

Content truncated to 500 words