Please save your time and energy on this question, a possible solution is provided.
I've found a book with two interesting indices. It lists Tamil proverbs and the book indexes all the initial words of the proverbs (the first index) and all the other words (the second index). How could we achieve that in TeX? I enclose an initial TeX file.
In this example we should get a
, Hello
and How
terms in the first index and all the other proverb terms in the second index. Please notice that symbols like commas, semi-colons etc. should not be a part of the indices.
% run: *latex mal-indexall-question.tex
\documentclass[a4paper]{article}
\usepackage[noautomatic]{imakeidx}
\indexsetup{firstpagestyle=empty}
\makeindex[name=initial, title=Initial words, columns=3]
\makeindex[name=noninitial, title=Non-initial words, columns=3]
\usepackage[colorlinks]{hyperref}
\begin{document}
\let\indexall=\index % To be changed...
Text before.\indexall{a," [b;] c: r: s: t; {d.}/ (e!) f? g| h¡ i¿}
Text in the middle.\indexall{Hello World!}
Text\indexall{How are you?} after.
\printindex[initial]
\printindex[noninitial]
\end{document}