Currently making extractions of messages that should contain some keywords, sometimes the number of occurrences retrieved is too big. The only way is either to redefine the keywords (HI% could become HIGH for instance) or to count the number of words between two keywords.