K-Means Subject Matter Expert Refined Topic Model Methodology
Abstract
We propose an innovative technique using K-means clustering to estimate the posterior topic distributions in Latent Dirichlet topic models as an alternative to the collapsed Gibbs sampling technique. This research also develops a topic modeling software instantiation of the K-means Subject Matter Expert Refined Topic methodology using the Visual Basic for Applications programming language. This topic modeling software is deployable across the majority of the Department of Defense computing environments and allows analysts to develop topic models using a graphical user interface.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 2017
- Accession Number
- AD1028777
Entities
People
- Nathan Parker
- Theodore T. Allen
- Zhenhuan Sui