Classifying Policy Issue Frame Bias in Philippine Online News

Document Type

Conference Proceeding

Publication Date

1-1-2023

Abstract

Media plays an important role in disseminating news to the public [1, 14]. However, the selection of what information to write about, how information is presented, and when it is broadcasted are within the control of the media outlet. Media can therefore shape the consumers’ opinion based on how publicly available news [1, 9] contents are published. Framing bias occurs when the author selects and highlights aspects of the news [8]. This study developed a model to classify the policy issue frames in select Philippine online news articles. Media Frame Corpus [5] was tested on the classification of policy issue frames using supervised learning methods including Bidirectional Encoder Representations from Transformers (BERT), Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), linear support vector machine (SVM), and logistic regression (LR). Results showed that the BERT model performed best with an accuracy of 74.73% with political frame (Frame 13) and economic frame (Frame 1) as leading policy issue frames. Implementing the MFC-BERT classification model on the Philippine dataset shows that there is a dominant policy issue frame across all selected media outlets. However, there is a significant difference in the usage of policy issue frames among these media outlets except for the Fairness and Equality frame (Frame 4) and the Quality of Life frame (Frame 10). Initially, ambiguous topics among media outlets when observed over time exhibit noticeable policy issue framing bias. However, a gradual transition to having similar policy issue frames among most media outlets occurred to these topics, similar to the unambiguous topics.

Link to Full Text

Share

COinS