Author-conference topic-connection model for academic network search

• DOI: 10.1145/2396761.2398597
• Publication Date: 2012-11-15T16:38:00Z
• Authors: Jianwen Wang; Xiaohua Hu; Xinhui Tu; Tingting He
• Keywords: Multinomial distribution

This paper proposes a novel topic model, Author-Conference Topic-Connection (ACTC) Model for academic network search. The ACTC extends the author-conference-topic (ACT) model by adding subject of conference and latent mapping information between subjects topics. It simultaneously models topical aspects papers, authors conferences with two layers: layer corresponding to topic, word topic. Each author would be associated multinomial distribution over (eg., KM, DB, IR CIKM 2012), conference(CIKM topics are respectively generated from sampled subject. Then words We conduct experiments on data set 8,523 authors, 22,487 papers 1,243 well-known Arnetminer website, train different number For qualitative evaluation, we compare three others LDA, Author-Topic (AT) ACT in search services. Experiments show that can effectively capture semantic connection types perform well expert searching searching.