ChronoSAGE: Diversifying Topic Modeling Chronologically
This paper provides an application of sparse additive generative models (SAGE) for temporal topic analysis. In our model, called ChronoSAGE, topic modeling results are diversified chronologically by using document timestamps. That is, word tokens are generated not only in a topic-specific manner, but also in a time-specific manner. We firstly compare ChronoSAGE with latent Dirichlet allocation (LDA) in terms of pointwise mutual information to show its practical effectiveness. We secondly give an example of time-differentiated topics, obtained by ChronoSAGE as word lists, to show its usefulness in trend detection.
KeywordsWord Pair Word List External Evaluation Word Token Word Probability
Unable to display preview. Download preview PDF.
- 2.Eisenstein, J., Ahmed, A., Xing, E.P.: Sparse additive generative models of text. In: ICML, pp. 1041–1048 (2011)Google Scholar
- 3.Griffiths, T.L., Steyvers, M.: Finding scientific topics. PNAS 101(suppl. 1), 5228–5235 (2004)Google Scholar
- 4.Newman, D., Karimi, S., Cavedon, L.: External evaluation of topic models. In: ADCS, pp. 11–18 (2009)Google Scholar