This repository contains code and data for an investigation into which of different notions of (Bayesian and information-theoretic) relevance best predict human judgements of an answer's helpfulness.
For detailed information about each subpart of the project, consult the README.md
file contained in each subdirectory.
analysis
contains all code related to data analysisexperiment
contains the code for the experimentplotting
contains code for additional plotting (on top of what is insideanalysis
)qualitative-analysis
contains code supporting the exploratory qualitative analysis on top ofanalysis
results
contains the data from the runs of the experiment