Let's be a bit more general and a bit more specific. I looked at data under DOS and saw it was a box :), so each line is either nearly all non-word chars (top/bottom), or one either side of content you want. So:

# clean lines with words in
$text =~ s/.*?             # anything
           (?:             # don't remember me
            (\w.*\w)       # everything from first to last word charac
+ter
                           # on the line - remember this
            |              # or
            [^\w\s]{6,}    # at least 6 characters that aren't words o
+r
                           # spaces (in a row) - forget me
           )
           .*              # the rest of the line
           /$1/gx;         # replace whole line with (\w.*\w) match,
                           # if made.
[download]

Now, if your text contains puntuation, you might want to amend the (\w.*\w) match to take account of commas, etc that may appear at the end of a line:

(\w.*\w) => ([\w\.,].*[\w\.,])
[download]

if text contained only commas and periods.

cLive ;-)

ps - IANAL (I am not a lecythis)