The app explores the transcripts of the Government Digital Service podcast, analysing the text for sentiment and characteristics
This data can be accessed via 3 tabs;
-
Sentiment tab - Details three plots, the first one shows both positive and negative sentiment per episode and the following two plots detail the sentiment average over the course of one & all episodes. The number of words to analyse average sentiment can be grouped by the word bin prompt e.g. If set to 30, for every 30 words the text will be analysed for positive and negative sentiment
-
Summary tab - Details a table of word counts and sentiment averages per episode
-
Characteristics tab - Details words that are characteristic of specific episodes, a good analysis to get the gist of what an episode is about
Treatment of data;
-
Stop words removed
-
Podcast hosts and guests removed
-
Sentiments with negation words like "not excited" have been reversed
In publishing the transcripts providing an open data source, The Government Digital Service Podcast
In developing and licensing for free the bing lexicon - Bing Liu and collaborators