CORD-19 NLP Project

Outline

Data Exploration

• Exploring the raw data

Dataset prep

• Filtering data within in MongoDB
• Creating the working dataset

Data Preprocessing for Text Analyses

• Further Text Preprocessing to prepare for modelling
• POS tagging and removal

Models

• Sentiment Analysis
• TD-IDF and LDA Models
• K-Means Clustering
• BERT
  • Embeddings
  • HDBSCAN
  • LDA Classification
  • TFIDF

Final Results and Comparisons

• Topics Derivation using TF-IDF from Bert HDBSCAN Clusters
• Topic Popularities Over Time
• Sentiment over time and by cluster
• Visualizations

Data Source

Data are from the meta data file for COVID-19 Open Research Dataset Challenge (CORD-19)