Every prior has two options which can be set, along with it’s particular inputs and options: -c for complement and -i for importance.
When the complement flag is on, the probability for a document that a prior generates will be the complement of its normal value. So for example, a text-search query which matches a document and would normally give a high score of perhaps 0.7, when the complement flag is on, would return a low score, like 0.3. The complement option essentially allows you to turn any prior into its opposite: you don’t want a document to contain a string, you don’t want documents authored by a particular author, you don’t want documents created near a certain time.
The importance option is for specifying how strong you want to assert a prior. p-search as a number of pre-defined importance levels: ‘none’, ‘low’, ‘medium’, ‘high’, ‘critical’, and ‘filter’. The importance level modifies the score a document would normally give. The following chat shows how the scores are modified:
Importance | Against | Supporting |
---|---|---|
Prior Score | 0.3 | 0.7 |
‘none’ | 0.5 | 0.5 |
‘low’ | 0.44 | 0.57 |
‘medium’ | 0.3 | 0.7 |
‘high’ | 0.11 | 0.91 |
‘critical’ | 0.001 | 0.999 |
‘filter’ | 0.0 | 1.0 |
Set the importance for a prior to determine how much you want it to affect the final score.