Is there a tool that allows me to extract bibtext entries matching a freely definable regular expression (e.g. all entries that do NOT contain the workd "example") without resorting to awk/grep etc.
biber has a new "tool" mode in the 1.5 dev version on SourceForge. With this, you can transform your .bib by using biber's reencoding and/or sourcemapping facilities as described in the manual. For example, to reencode your UTF-8 .bib and replace all UTF-8 with LaTeX macros:
biber --tool --output_encoding=ascii --output_safechars file.bib
To address the sorts of transformations you need, create a biber config file with a sourcemap section as per the manual. You can fairly arbitrarily transform the entries, including with regular expression search/replace, filter entries, drop/merge fields etc.
Bear in mind that biber 1.5 DEV requires biblatex 2.5 DEV from SourceForge for normal "non tool mode" use.