Extraction of the financial policy topics by latent dirichlet allocation

We conducted topic extraction concerning the financial policy topics in Japan just after the East-Japan great earthquake disaster in 2011. The target document is text data of the Policy Board of the Bank of Japan financial policy meeting proceedings summaries. The topic extraction methods we used was the LDA algorithm. We extracted interesting three topics from the summaries. One topic has the peaks which overlapped the Japanese stock price fall times. The topic is considered to be related to the risk around the future of Japan. We think that another disaster-related topic expresses the uncertainty of the company production activity. In this paper, we used other economics time series data as plausible reasons to explain the extract topic. We discuss in the paper our approach by economics data to give a plausible excuse for the topic.