One way you could do it, is by removing the 01 and ! and just putting each message on a new line, then use the tokenat system expression with the delimiter being the newline thing (character?). Or, you could use a regex, in a general case, this regex would work: "[\d]+\s([\w\s]+\!)". For the purpose of extracting a string, if you know the number of the text you want, have that number stored, as a variable or parameter and then this regex would work: number & '\s([\w\s]+\!)'".
So this: would always return the first regex match (text):
However, this would allow you to specify the number: