Abstract
The questions on the ethical codes of journalism is as old as the occupation itself. The questions which reveal themselves as objectivity/subjectivity differentiation are re-addressed in digital era. Questioning to what extent news articles should involve self-opinions determine journalism because journalists themselves defined their notion of subjectivity/objectivity by using philosophical and scientific thoughts or methods. Additionally, the discussion from the perspective of society is on whether can be maintained a democratic public sphere and having reliable news sources or not. Our motivation is to develop an Information System that detects news at the digital era. Digital era enters into age of big data that the number of news articles published is beyond editorial boundaries or already prepared by an AI. In this work, we measure sentence-wise objectivity/subjectivity with machine learning methods by using News Subjectivity Dataset. Firstly, the sentence features are extracted with Language Model and fixed length features used for training SVM classifier. The classifier performance has .69 accuracy. The results show that although we are at the beginning of measuring news biases, the accuracy gives an optimistic view of the field.
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
KEYWORDS
Neural Network, Language Model, Support Vector Machines, Habermas, Public Sphere