Topic segmentation model based on ATNLDA and co-occurrence theory and its application in stem cell field. (June 2013)

• Record Type: Journal Article
• Authors: Wu, QingQiang; Zhang, CaiDong; An, XinYing
• Is Part Of: Journal of information science. Volume 39:Number 3(2013)
• Page Start: 319
• Page End: 332
• Publication Date: 2013-06
• Subjects: automatic topic number LDA (ATNLDA) -- correlation analysis -- co-occurrence theory -- latent Dirichlet allocation (LDA); Information science -- Periodicals; Information science; Periodicals; 020.5
• DOI: 10.1177/0165551512457893
• Languages: English
• ISSNs: 0165-5515

Abstract

This paper describes the application of co-occurrence and latent Dirichlet allocation (LDA)-based topic analyses in stem cell-related literature research. On account of the deficiency of parameter estimation in LDA, this study integrated co-occurrence theory and clustering judgement indicators and constructed an ATNLDA (Auto Topic Number LDA) model for topic segmentation. Next, ATNLDA was used to determine the optimal topic number of stem cell research literatures from 2006 to 2011 in PubMed, which was then used for topic segmentation of research content in stem cell data set. After stem cell research topics were obtained, they were analysed in terms of topic label, topic research content and interrelation between topics. The results verified that application of ATNLDA in topic segmentation in stem cell literature research is effective and feasible. Current deficiencies of ATNLDA and future study plan were also discussed.