Getting a documents probability vector when using QL retrieval

[noamaon]

Is there a way to access the term distributions created when using a Bayesian smoothed language model on indexed documents? If I understand correctly, these distributions should be created when running search with the --qld argument.