There are several query strategies available.
Parameters should be under the section
As it says: randomly select samples with no regard to model assigned probabilities. Warning: selecting this option means your review is not going to be accelerated by ASReview.
Choose the most uncertain samples according to the model (i.e. closest to 0.5 probability). Doesn’t work very well in the case of LSTM’s, since the probabilities are rather arbitrary.
Choose the most likely samples to be included according to the model.
Use clustering after feature extraction on the dataset. Then the highest probabilities within random clusters are sampled.
A mix of two query strategies is used. For example mixing max and random sampling with a mix ratio of 0.95 would mean that at each query 95% of the instances would be sampled with the max query strategy after which the remaining 5% would be sampled with the random query strategy. It would be called the max_random query strategy. Every combination of primitive query strategy is possible.