Skip to main navigation Skip to search Skip to main content

User profiling by combining topic modeling and pointwise mutual information (TM-PMI)

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

5 Scopus citations

Abstract

User profiling is one of the key issues in personalized recommendation systems. A content curation social network is a content-centric network; it encourages users to repin items from other users and other websites. It further permits users to arrange the pins according to their interests. It is therefore possible to estimate user interest from the pins. In this paper, we propose a user profiling approach to combining topic model and pointwise mutual information (TM-PMI). We first extract a pin’s description, and then apply latent Dirichlet allocation (LDA, one of the topic modeling schemes). A three-layer hierarchical Bayesian model of user-topic-word is thus obtained. Then, a personal model is obtained by selecting a set of correlated words with constraints of word probability and PMI. The experimental results confirm the efficiency of the proposed approach.

Original languageEnglish
Title of host publicationMultiMedia Modeling - 22nd International Conference, MMM 2016, Proceedings
EditorsRichang Hong, Nicu Sebe, Qi Tian, Guo-Jun Qi, Benoit Huet, Xueliang Liu
PublisherSpringer Verlag
Pages152-161
Number of pages10
ISBN (Print)9783319276731
DOIs
StatePublished - 2016
Event22nd International Conference on MultiMedia Modeling, MMM 2016 - Miami, United States
Duration: Jan 4 2016Jan 6 2016

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9517

Conference

Conference22nd International Conference on MultiMedia Modeling, MMM 2016
Country/TerritoryUnited States
CityMiami
Period01/4/1601/6/16

Keywords

  • Latent dirichlet allocation
  • Pointwise mutual information
  • Topic modeling
  • User profile

Fingerprint

Dive into the research topics of 'User profiling by combining topic modeling and pointwise mutual information (TM-PMI)'. Together they form a unique fingerprint.

Cite this