Does IPscreener adapt to rankings done in previous versions, e.g. match or noise to exclude such results that are not relevant?

Yes, the matching of new version is optimized according to rankings done in a case. Rating a document as "match" means the technology therein is almost identical to the query, a "relevant" rank indicates a document with a relevant solution partly overlapping the query, a "background" label refer to a general solution within the technical field and and "noise" indicates areas that should be excluded in coming matching procedures.