Biber's Variation across Speech and Writing (Cambridge: CUP, 1988) considers the following as hedge uses (the POS tags are as per UCREL CLAWS C7 tagset). If your corpus is POS tagged, it should be easy to extract these patterns.
about_R*
something_PN1 like_II
more_R* or_CC less_R*
almost_RR
maybe_RR
sort_RR21 of_RR22
kind_RR21 of_RR22