I think this is a case where a single big(gish) regex is not the best way to extract the information. A combination of split and grep with simple regexes seems a better choice.

First split the text on either two or more blanks, or a line feed. That isolates the candidate phrases mixed with other text fragments. Then select three-word-phrases by keeping only those fragments that contain exactly two blanks. Finally, select for the word "NOTICE". Put together:

print "$_\n" for
    grep /\bNOTICE\b/, grep tr/ // == 2, split / {2,}|\n/, $text;
[download]

Anno