12:40 PM - 1:00 PM
[4A2-J-3-03] A Study on Topic Transition Analysis Method Using FKC Corpus
Keywords:Transition Analysis, Neural Network, FKC Corpus, LDA, Markov Chain
With the spread of SNS, text mining to acquire useful knowledge from enormous data has been active. Moreover, much attention has been paid to time series data and many studies predicting social phenomena have been reported. However, there are few studies focusing on the seriesiness of text data for each user and it can be expected to be a new viewpoint of data analysis in near future.
In this paper, we discuss the behavior modeling of users based on text data. We propose a method using LDA as the document representation and Neural Network as the transition modeling. By using Neural Network, it becomes possible to expand the expression ability and flexibility, and then, the improvement of modeling can be expected compared with the conventional method. In fact, we carry out an experiment and report that the performance of the proposed method is improved. We also show an example of topic transition analysis and mention the practicality of the method.
In this paper, we discuss the behavior modeling of users based on text data. We propose a method using LDA as the document representation and Neural Network as the transition modeling. By using Neural Network, it becomes possible to expand the expression ability and flexibility, and then, the improvement of modeling can be expected compared with the conventional method. In fact, we carry out an experiment and report that the performance of the proposed method is improved. We also show an example of topic transition analysis and mention the practicality of the method.