Tasks

[matthewr6]

• Biterm model finalizing + experiments (Chris)
• TKM experiments (Matthew)
• K-means experiments (different # clusters) (Sasankh)
• LDA experiments (different featurizers/clusters) (Matthew)
• Pseudo tf-idf topic extractor (based on clusters rather than documents - tf within cluster, idf against other clusters - Chris
• Centroid vector extractor (Matthew)
• MultinomialNB max probability-based extractor (Sasankh)
• Brief descriptions/writeups of each featurizer, model (+ extractor), and experiment (if experiment is interesting; from results and visualizations)
• Poster (Sasankh + Chris)
• Start on paper (Chris + Sasankh)

Graphs (Matthew - dependent on data from experiments)

• accuracy differences
• for most successful model - plot of clusterings
  • cluster topics
  • examples of articles

[matthewr6]

update on TKM model - very poor clustering (see PCA graphs), decided to drop it

update on centroid vector extractor - done

• write up the featurized-text vs. keyword-text analysis method (Matthew)
• clustering analysis per Chuma's suggestion (Sasankh or Chris?)
• qualitative analysis on graphs and cluster.json files (Sasankh or Chris?)