Clark Kent at SemEval-2019 Task 4: Stylometric Insights into Hyperpartisan News Detection

Proceedings of the 13th International Workshop on Semantic Evaluation (SemEval-2019), pages 934–938 Minneapolis, Minnesota, USA, June 6–7, 2019. ©2019 Association for Computational Linguistics 934 Clark Kent at SemEval-2019 Task 4: Stylometric Insights into Hyperpartisan News Detection Viresh Gupta, Baani Leen Kaur Jolly, Ramneek Kaur, Tanmoy Chakraborty Indraprastha Institute of Information Technology, Delhi (IIIT-Delhi), India {viresh16118, baani16234, ramneekk, tanmoy}@iiitd.ac.in Abstract In this paper, we present a news bias predic- tion system, which we developed as part of a SemEval 2019 task. We developed an XG- Boost based system which uses character and word level n-gram features represented using TF-IDF, count vector based correlation matrix, and predicts if an input news article is a hyper- partisan news article. Our model was able to achieve a precision of 68.3% on the test set provided by the contest organizers. We also run our model on the Buz- zFeed corpus and find XGBoost with simple character level N-Gram embeddings to be per- forming well with an accuracy of around 96%. 1 Introduction The problem of hyperpartisan news detection (Pot- thast et al., 2018) is based on predicting whether a news article is biased towards a specific political wing or not. The problem falls under the category of classification problems, and the task is to clas- sify an article as being extremely one-sided or not. A closely related problem is that of fake new de- tection wherein the task is to analyze the veracity of an article, and classify it based on some prede- fined degrees of truthfulness. Our problem has a high societal relevance, since one-sided news poses a great threat to democracy, particularly in the context of conduct- ing fair elections. In this paper, we discuss our approach to solving this problem used during the contest Hyper Partisan News Detection, a com- petition task at SemEval 2019 (Kiesel et al., 2019). More formally, our problem definition is: Definition 1 (Hyperpartisan News Detection) We are given a set of news articles A, where each article ai is marked with two labels: a Boolean label hyperpartisan hi which indicates if article ai is biased towards a political wing, and a bias label bi ∈ {left, right, left-center, right-center, least} which indicates which wing the article is biased towards. If hi = True, then bi ∈ {left, right}; if hi = False, then bi ∈ {least, left-center, right-center}. The objective is to learn a classifier C which predicts the hyperpartisan label hj for an unknown news article aj. In this work, we identify the role of various tra- ditional NLP features in determining the degree of partisanship. We utilise standard term-frequency and inverse document frequency vector features computed for uni, bi and tri-grams obtained from the corpus. We do this feature extraction at both character and word level and then train a gradi- ent boosted decision tree as a classifier for identi- fying partisanship. We also compare other meth- ods of classification such as SVM, KNN, Naive Bayes and Logistic Regression for the task using the same …

Content truncated to 500 words