CS590/690 SEMINAR
A Neural Network’s Insight into Predicting Trending Articles in News Streams
Pouya Ghahramanian
Ph.D Student
(Supervisor: Prof.Dr. Fazlı Can )
Computer Engineering Department
Bilkent University
Abstract: In this research, we have two primary objectives. First, we detail our methodology for compiling a comprehensive dataset, comprised of articles and associated user comments from The New York Times spanning the last 10 years. This dataset reveals the evolving nature of news and reader interactions over the decade. Subsequently, we address the challenge of predicting which news articles will trend in real-time news streams. Feedback from readers, reflected as comments or upvotes, serves as a metric to identify articles receiving substantial attention. However, the criteria for “trending” can shift due to changes in reader interests or significant global events. This variability is referred to as concept drift. We introduce three innovative neural network models tailored to accommodate these drifts. Additionally, we discuss the adaptive mechanisms of these models as they continuously learn from incoming articles without the need for extensive retraining. Preliminary experiments conducted using our new models, compared with some existing baselines across three news streams, demonstrate the potential effectiveness of our approaches
DATE: October 9th, Monday @ 15:50 EA-502