Bag of Timestamps: A Simple and Efficient Bayesian Chronological Mining

Authors:
Tomonari Masada;Atsuhiro Takasu;Tsuyoshi Hamada;Yuichiro Shibata;Kiyoshi Oguri
Affiliations:
Nagasaki University, Nagasaki, Japan;National Institute of Informatics, Tokyo, Japan;Nagasaki University, Nagasaki, Japan;Nagasaki University, Nagasaki, Japan;Nagasaki University, Nagasaki, Japan
Venue:
APWeb/WAIM '09 Proceedings of the Joint International Conferences on Advances in Data and Web Management
Year:
2009

Citing 5
Cited 0

Latent dirichlet allocation

The Journal of Machine Learning Research
Group and topic discovery from relations and text

Proceedings of the 3rd international workshop on Link discovery
Dynamic topic models

ICML '06 Proceedings of the 23rd international conference on Machine learning
Topics over time: a non-Markov continuous-time model of topical trends

Proceedings of the 12th ACM SIGKDD international conference on Knowledge discovery and data mining
Multiscale topic tomography

Proceedings of the 13th ACM SIGKDD international conference on Knowledge discovery and data mining

Quantified Score

Hi-index	0.01

Visualization

Abstract

In this paper, we propose a new probabilistic model, Bag of Timestamps (BoT) , for chronological text mining. BoT is an extension of latent Dirichlet allocation (LDA), and has two remarkable features when compared with a previously proposed Topics over Time (ToT) , which is also an extension of LDA. First, we can avoid overfitting to temporal data, because temporal data are modeled in a Bayesian manner similar to word frequencies. Second, BoT has a conditional probability where no functions requiring time-consuming computations appear. The experiments using newswire documents show that BoT achieves more moderate fitting to temporal data in shorter execution time than ToT.