Especially evaluate period of the random walks in terms of:
- Quality of the results
- Performance Gain
Potential Periods:
- 12 weeks (current)
- 4 weeks
- shorter/longer?
Check if a shorter period for random walks would allow on-demand runs of the whole model, rather than weekly calculations of the policy terms. Therefore:
- Check performance of / time needed to run the whole workflow (including tokenization of new tweets!)
- Evaluate if adding more recent tweets / the tweets to be evaluated themselves into the model yields improved results
Potentially adjust time periods for Seed Term generation. However, this seems sufficiently fast
Especially evaluate period of the random walks in terms of:
Potential Periods:
Check if a shorter period for random walks would allow on-demand runs of the whole model, rather than weekly calculations of the policy terms. Therefore:
Potentially adjust time periods for Seed Term generation. However, this seems sufficiently fast