Abstract | ||
---|---|---|
Predicting popularity of a post in microblogging services such as Twitter is an important task beneficial for both publishers and regulators. Traditionally, the prediction is done through various manually designed features extracted from post and user contexts. In recent years, deep learning models such as convolutional neural networks (CNN) have shown significant effectiveness in image processing. In this paper, we make a novel investigation of the effectiveness of deep learning models in predicting image post popularity, with the raw image as the input. In contrast to previous works that use existing model trained for object detection, we trained a CNN model targeting directly at predicting popularity. We show that a dedicated CNN is more effective than networks trained for other purposes and is comparable to text-based predictors. |
Year | DOI | Venue |
---|---|---|
2019 | 10.1007/978-3-030-15712-8_56 | european conference on information retrieval |
Field | DocType | Citations |
Object detection,Data mining,Social media,Convolutional neural network,Computer science,Popularity,Microblogging,Image processing,Artificial intelligence,Deep learning,Machine learning | Conference | 2 |
PageRank | References | Authors |
0.39 | 0 | 2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Yihong Zhang | 1 | 9 | 10.65 |
Adam Jatowt | 2 | 903 | 106.73 |