This document summarizes a conference paper on a dataset of online news articles from the Mashable website between 2013-2015. It contains over 39,000 data points with 61 attributes like number of shares, topics, sentiment, and keywords. The authors analyzed which attributes predicted popularity, defined as over 1,400 shares. They found publication day, channel/topic, and LDA topic distributions were most correlated with shares. Random forest models using these attributes achieved an AUC of 0.73 for predicting popularity. The authors noted challenges in understanding the data and modeling its relationships, and recommended ways to improve data collection and article publishing strategies.