I have a procedural question about how relevance language works in action:
In a single report, is all relevance to be processed aggregated and de-duplicated to reduce processing of identical queries, or is each query processed separately (i.e. two identical fixlets with identical relevance results in double the relevance processing on the evaluating client).
Is relevance only processed in complete statements, or is it processed in reusable segments, such as say, “name of operating system” processed and can be reused in longer statements like “name of operating system contains ‘win2’”? It seems to be only complete statements, I’d just like confirmation.
Any other specific details (whether high level or detailed technical) about how relevance is actually processed would be appreciated. For reference, I’m only concerned about 9.5+, so if older versions processed differently, that’s not relevant 