Search Personalization in Folksonomy by Exploiting Multiple and Temporal Aspects of User Profiles
Lists
Only you see this
Abstract
Social tagging data, also known as folksonomy, are a valuable indication for the user’s understanding of a resource. The nature of folksonomy data in which a user annotates a resource with their opinions provides immense potential to contribute to search personalization. The challenge lies in extracting interests from the folksonomy data and building accurate user profiles while maintaining their characteristics. Furthermore, the current state-of-the-art technologies that utilize folksonomy for search personalization have not fully exploited both multiple and temporal aspects in user profiles. In this paper, we propose a search personalization framework that constructs a user profile network with identification of the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">multiple topics</i> of the user and the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">temporal values</i> of tags. Then, the user profile network is further explored through a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">link analysis technique for the network</i> to score the tags by their importance. The performance of the proposed framework is evaluated against various state-of-the-art folksonomy-based personalization models and it consistently outperforms all of the compared models under the conditions of the best combination of ranking functions and link analysis techniques.
BibTeX
@inproceedings{han2019search,
author = {Keejun Han and Mun Yong Yi and Jungeun Kim},
title = {{Search Personalization in Folksonomy by Exploiting Multiple and Temporal Aspects of User Profiles}},
booktitle = {IEEE Access},
year = {2019},
url = {https://openreview.net/forum?id=gKB-TybtC-G},
}Loading related content...