Study Topic Parameters

Study topic parameters

Created: 2022-09-06 14:21
#note

Investigate how language affects results
multiple plots for the datasets (one figure for datasets, composed by multiple plots):
x-axis techniques
y-axis value of a score

OR (one figure for score, composed by multiple datasets)

x-axis technique
y-axis

TO-DOs:

• Investigate number of topics (elbow rule)
• Investigate parameters
• plot as explained above
• see literature to check if our results are valid compared to the ones in the papers
• discrepancy on number of topics (between models and datasets) -> find explaination
• draft section 2 of the paper

Questions:

• For those methods that can decide automatically the optimal number of topics, do we have to compute the coherence and diversity with such number of clusters?
• To compute the score metrics, how many words do we want to consider (top10, top20)?

References

1. How to determine optimal number of topcis

Code

1. My code

Tags