SFI Complex System Summer School project
Aim: explore the dynamics of natural language conversations into opinion space.
New Steps:
- Translate speeches to English
- Extract relevant speeches using keywords
- Get exploratory plots
- Number of speeches vs time (by party, by politician)
- Length of these speeches
- Get some basic history understanding
- Landmark votes
- Key parties/players
- Important news stories
- Hand-label some relevant speeches to make validation set (10 per decade?)
- Improve model
- Include wikipedia-neutral corpus into model data
- Train models (basic ones first)
- Test on validation data of speeches
- If good: Run model on whole dataset