Clean the source before designing
Work on a copy of the original text. Remove navigation, captions, addresses and repeated headings. List meaningful terms, check spelling and resolve inconsistent naming. Decide whether forms such as ‘project’ and ‘projects’ belong together. Merge terms by meaning rather than appearance, and do not remove a negative if it changes what a response says.
Frequency and importance are different
In product launch notes, support, deadlines and quality may occur different numbers of times. If you show frequency, count occurrences using a consistent rule and keep the counts separately. If you create an editorial summary, select the key ideas manually and label that choice. A cloud cannot tell whether someone praised quality or complained about it. Return to the surrounding sentences before interpreting a prominent term.