Analyzing Inaugural Addresses with NLP

Check out my blog post (TBU) .

Description

Utilizing topic modelling and dimensionality reduction, I analyze topic trends through time and cluster presidents.

Data

• OSF , includes all addresses excluding President Biden
• Pasted President Biden's address into a text file

File Contents

• Data/
  • InauguralTexts/ contains text files of all addresses
  • PresidentInfo/Pres_Details.csv contains further details on the presidents such as, political party, start date, end date
  • Visualization contains images utilized in my presentation and blog posts
• Code/
  • 01_Data_Cleaning.ipynb contains all data aggregation and cleaning necessary for this project
  • 02_EDA_Analysis.ipynb contains all initial EDA (i.e. word counts, words per sentence, "I" vs. "We")
  • 03_NLP_Analysis_Aggregate_Tokenized.ipynb contains initial topic modelling (LDA/NMF) and clustering at the address level.
  • 04_NLP_Sentence&Par_Tokenized.ipynb contains NMF topic modelling at both the sentence and paragraph level. Topic vectors were averaged for each president. These results are the bulk of my presentation and blog post.
  • 05_CorEx.ipynb contains topic modelling with CorEx. Seeding initial topics with CorEx did not product results as well as NMF topic modelling at the paragraph level and so was not discussed in my presentation or blog post.
  • 06_Sentiment_Analysis.ipynb contains sentiment analysis utilizing nltk's VADER
  • Classes_Functions.py contains functions used throughout notebooks 01-06