As I read the papper "[Optimizing Top-k Document Retrieval Strategies for
Block-Max Indexes](http://engineering.nyu.edu/~suel/papers/bmm.pdf)", i see that each block of posting list stores one max score using to skip the block whose max score is less than then θ value,and max score value is based on term frequency and document length.
My question is that can we use the score stored in payload as max score? I have a slow query as follows:
GET my_index/_search
{
"query": {
"bool": {
"must": [
{
"terms": {
"tags": [
"101",
"102",
"103",
...
]
}
}
]
}
}
}
Hundreds of terms in tags list , this is very slow , because es need to merge hundreds of posting list , my idea is to store a quality score in payload , and then use the quality score as max score of block, so this maybe can skip lots of block whose quality score not in Top-K hits
Does this idea es support?
OR can block max wand algorithm use the value of Index Sorting field as score?