Topic Extraction Analysis for Monetary Policy Minutes of Japan in 2014

A Text Mining Analysis of Central Bank Monetary Policy Communication in Nigeria

“…This sort of probabilistic generative process is then used to determine the probabilities of a word and a document belonging to a topic 7 . Following the procedure outlined in Calvo-González et al, (2018) and Shirota et al (2015), the LDA generative process for our corpus is described in the following steps:…”

“…The FOMC's communication strategy involves issuing a policy statement after each meeting while the minutes of the meetings are released to the public three weeks after the policy decision date. Similar monetary policy communication strategies are in place in several other central banks around the world, including the Bank of England, Bank of Japan, Bank of In line with the increased level of communication among central banks, a new but growing body of literature has focused on the extraction of useful insights from monetary policy documents using different text mining techniques (see Shirota, 2015;Bruno, 2016;Kahveci and Odabas, 2016;Luangaram and Wongwachara, 2017;Oshima and Matsubayashi, 2018;Park et al, 2019;Shapiro and Wilson, 2019;Omotosho, 2020). According to Blinder et al (2008), different methods are available for the analysis of central bank communication; including the indirect approach (whereby financial market reactions to policy announcements are measured using event windows), and a direct approach (whereby the researcher provides a subjective assessment of central bank communication by manually reading through the policy document and coding its contents into numerical scales).…”

Central Bank Communication in Ghana: Insights from a Text Mining Analysis

“…We employ the Latent Dirichlet Allocation (LDA) modelling approach developed by Blei et al (2003) to cluster the texts in our corpus into meaningful categories that describe the main topics in the BOG's monetary policy press releases. Based on the procedure outlined in Calvo-González et al, (2018) and Shirota et al (2015), the LDA generative process for our corpus is described as follows:…”

“…Thus, we follow the practice in literature by using the collapsed Gibbs sampling algorithm to approximate the posterior distributions of the hidden variables given in equation 3by running 5,000 iterations. As noted by Shirota et al (2015), the collapsed Gibbs sampling algorithm is a useful procedure for approximating posterior distributions that are difficult to compute directly. Once the posterior estimates for φ and θ are derived, the algorithm then returns the topic representation of each document in the corpus.…”

“…The increased level of central bank communication as well as the availability of relevant automated algorithms have enabled the growth of a strand of research focused on the use of text mining to extract useful insights from monetary policy documents (see Park et al, 2019;Shapiro and Wilson, 2019;Oshima and Matsubayashi, 2018;Luangaram and Wongwachara, 2017;and Shirota, 2015). While the communication strategies of central banks in developed countries have been analysed (see Siklos, 2013;Shapiro and Wilson, 2019;Hubert and Fabien, 2017;Kahveci and Odabas, 2016;Luangaram and Sethapramote, 2016;Binette and Tchebotarev, 2019), the literature for central banks in developing countries (including Ghana) is still relatively scarce.…”