Presentation feedback

[bensoltoff]

• Any way to filter tweets by user location? Not geocoding, but in their profile they can identify where they live. Look for Venezuela cities or country name?
• Any native Spanish sentiment dictionaries? Not necessarily in tidytext, but could potentially import if you have doubts about the translation accuracy
• Remove search terms as stopwords
  • Or instead of raw term frequency, consider tf-idf
• For presentations, scale ggplot2 graphs up using theme_*(base_size = 22) or a larger value
  • Also consider an R Markdown presentation format, rather than the raw R Presentation built in to RStudio
• Lots of diagnostic-type graphs shown here. What is the final presentation format? A research paper? Figure out which are the most important graphs. What tells the story the best? Right now it feels like chart overload