Paper Info
Title
Improving the Gain of Visual Perceptual Behaviour on Topic Modeling for Text Recommendation.
Abstract
Internet information services have been greatly improved profiting from the growing performance of interest mining technology. Visual perceptual behaviours, a new hotspot of mining user's interests, have resulted in great gains in some typical Internet information services, e.g., information retrieval and recommendation. It is validated that combining the subjective visual perceptual behaviours with the objective contents can significantly improve these services' performance. However, the existing methods usually treat the contents and visual perceptual behaviours as two independent parts in the calculating process. The gain of visual perceptual behaviours has not been fully exploited. In this paper, we mainly aim at improving the gain of visual perceptual behaviour for text recommendation, by integrating the objective contents with subjective visual perceptual behaviours. We investigate the correlation between user's reading interests and records of real-time interaction on texts, and then design a real-time visual perceptual behaviour based method for text recommendation, which is able to: (1) build a joint interest model, called ViP-LDA (Visual Perceptual LDA), by integrating the user's visual perceptual behaviours into topic model; (2) make more accurate text recommendation based on ViP-LDA with feedback adjustment. Several experiments on a real data set are implemented to demonstrate the effectiveness of our method.
Year
DOI
Venue
2017
10.1145/3132847.3133122
CIKM
Keywords
Field
DocType
Visual Perceptual Behaviour, Eye Tracking, Text Recommendation, LDA, Interest Model
Internet information services,Information retrieval,Computer science,Eye tracking,Topic model,Hotspot (Wi-Fi),Visual perception
Conference
ISBN
Citations 
PageRank 
978-1-4503-4918-5
0
0.34
References 
Authors
5
4
Name
Order
Citations
PageRank
Cheng Wang15811.05
Yujuan Fang200.34
Zheng Tan300.34
Yuan He460.78