Tagging Items Automatically Based on Both Content Information and Browsing Behaviors.

Shen Liu,Hongyan Liu

DOI: https://doi.org/10.1287/ijoc.2020.1007

IF: 3.288

2021-01-01

INFORMS Journal on Computing

Abstract:Tags have been adopted by many online services as a method to manage their online resources. Effective tagging benefits both users and firms. In real applications providing a user tagging mechanism, only a small portion of tags are usually provided by users. Therefore, an automatic tagging method, which can assign tags to different items automatically, is urgently needed. Previous works on automatic tagging focus on exploring the tagging behavior of users or the content information of items. In online service platforms, users frequently browse items related to their interests, which implies users’ judgment about the underlying features of items and is helpful for automatic tagging. Browsing-behavior records are much more plentiful compared with tagging behavior and easy to collect. However, existing studies about automatic tagging ignore this kind of information. To properly integrate both browsing behaviors and content information for automatic tagging, we propose a novel probabilistic graphical model and develop a new algorithm for the model parameter inference. We conduct thorough experiments on a real-world data set to evaluate and analyze the performance of our proposed method. The experimental results demonstrate that our approach achieves better performance than state-of-the-art automatic tagging methods. Summary of Contribution. In this paper, we study how to automatically assign tags to items in an e-commerce background. Our study is about how to perform item tagging for e-commerce and other online service providers so that consumers can easily find what they need and firms can manage their resources effectively. Specifically, we study if consumer browsing behavior can be utilized to perform the tagging task automatically, which can save efforts of both firms and consumers. Additionally, we transform the problem into how to find the most proper tags for items and propose a novel probabilistic graphical model to model the generation process of tags. Finally, we develop a variational inference algorithm to learn the model parameters, and the model shows superior performance over competing benchmark models. We believe this study contributes to machine learning techniques.