A Deep Learning-Based Sentiment Classification for Identifying Advertorial Content in Online News

Abstract

The rapid advancement of technology and the widespread use of the internet have brought significant positive and transformative impacts across various aspects of human life, including finance, healthcare, education, and the media industry. One notable consequence of information transparency is the vast availability and large-scale exchange of data. However, this also presents new challenges, particularly in the spread of misleading content such as disguised advertorials that resemble genuine news. This threatens the objectivity of the information received by the public. To address this issue, an automated solution is needed to identify the distinguishing characteristics of advertorials in online news content. This study proposes a deep learning approach using the Convolutional Neural Network (CNN) model to detect sentiment as an indicator of advertorial content. CNN is a widely used deep learning model for processing sequential and spatial data, capable of automatically learning features from text. The dataset comprises news articles categorized by advertorial traits, such as positive or neutral sentiment, persuasive language, and promotional content highlighting specific entities. The data undergo several processing stages, including text preprocessing, tokenization, padding, and CNN model training. Model performance is evaluated using accuracy, precision, recall, and F1-score. The experimental results show a validation accuracy of 84%, although overfitting issues were observed. Despite ongoing limitations, such as restricted data and suboptimal parameter tuning, the findings suggest that the CNN model has potential for automatically detecting advertorial content and can serve as a basis for future research using more advanced models and refined parameter adjustments.