Fedspeak - 2020 Shiny Contest Submission

Fedspeak

Authors: Jay Ulfelder
Working with Shiny for 1+ years

Abstract: This dashboard applies common natural language processing (NLP) routines to public speeches by U.S. Federal Reserve officials since 2006. For followers of the Fed, the dashboard offers a quick way to get a read on the sentiment and topics of new Fed speeches and how they compare to past examples. For a wider audience, the dashboard demonstrates some of the possibilities and power of text mining.

Full Description: Fedspeak applies common natural language processing (NLP) routines to public speeches by U.S. Federal Reserve officials since 2006, as scraped from the Fed's online archive.

This app serves two purposes. For followers of the Fed, the dashboard offers a quick way to get a read on the sentiment and topics of new Fed speeches and how they compare to past examples. For a wider audience, the dashboard demonstrates some of the possibilities and power of text mining.

Users interact with the tool through several tabs.

  • The Pick a Speech tab serves as the app's landing page. Here, you can select one of the more than 800 speeches by Fed officials since 2006 and read the full text, as scraped from the Fed's online archive. The display defaults to the most recent speech.

  • The Sentiments tab uses gauges and column charts to quantify the sentiment of the selected speech and compare it to all speeches from 2006 to early 2020.

  • On the Topics tab, you can see how the speech gets classified by a set of four topic models derived from the historical archive. Gauges show the probability that the speech belongs to each one of the four topics, and bar charts show the terms associated with those topics.

  • The Historical trends tab uses dot plots with loess curves and a stream graph to show how the distribution of sentiment scores and topics has changed over time.


Category: Research
Keywords: nlp, text mining, sentiment analysis, topic modeling, economics, political economy
Shiny app: https://dartthrowingchimp.shinyapps.io/fedspeak/
Repo: https://github.com/ulfelder/practice-projects/tree/master/fedspeak
RStudio Cloud: https://rstudio.cloud/project/1009663

Thumbnail:
image

Full image:

2 Likes