show opposite of "more like this" in elasticsearch -

i'm using elasticsearch pull logs. when use api search (say, "where log_status=error"), lots of results. i'd show nice smattering of entries- opposite of "show more this".

cardinality gives great aggregation of returned results, doesn't score every result.

what i'm after more complicated fuzzy search; want know uniqueness of documents each other, not accuracy of match query. in fact, current _scores 1.0 because actual query *.

it's okay if solution expensive (say, script or function_score query). don't know how it.. or if out of scope.

i don't see way in es. sounds might need cluster documents prior loading them es. query entry point essentially, , want see pairwise similarity between returned docs... correct? if so, complicated, because if use kmeans or something, know docs in same cluster, might not score between them. may consider using kmeans dimensionality reduction mechanism pairwise similarity processing.

if don't have millions of docs, may able generate score between docs client side working response es. if have ton of docs (millions) you've got tougher problem, , might need use powerful cluster , compare prior indexing. also, consider getting docs es, , doing mlt using each doc more docs... score between docs... i;m thinking out loud, depends on how volume have.