The basic answer to your question is that I don't think it's possible to achieve your sorting objective in Word.
According to Word 2000's help facility, Word uses the following rules:
“If you sort by text: Microsoft Word first sorts items that begin with punctuation marks or symbols (such as !, #, $, %, or &). Items that begin with numbers are sorted next; and items that begin with letters are sorted last. Keep in mind that Word treats dates and numbers as though they were text. For example, "Item 12" is listed before "Item 2."
If you sort by numbers: Word ignores all characters except numbers. The numbers can be in any location in a paragraph.”
Now I have had similar problems with sorting in the past, and I have tried various work-arounds such as temporarily using find-and-replace to substitute a punctuation character for the space between two words. The problem with this is that Word does not simply overlook the punctuation character and treat the item consisting of word-punctuation char.-word as a single unit. It alphabetises the punctuation character as well, which negates the point of the substitution.
I think that especially in your case, a better bet for achieving your end would be to use a database program that allows you to specify your sort criteria more flexibly than Word allows.
That might seem like more work because you would have to create a new database entry for each term, but I think it would ultimately be worth it for your purpose.
The reason is that if you structure it sensibly you will be able to break up each entry into all its components, such as headword, definite/indefinite article in headword (if relevant), date of first usage, country/countries where current, primary definition, secondary definition (etc.), source of data (e.g. name/date of a TV show, rap record, magazine, novel etc. in which a term is believed to have been used for the first time), first example of usage, second example of usage (etc.), offensiveness marker (if relevant/desired), and any other types of data item you might decide to include.
This will both enable you to analyse your data in a way that you could not do otherwise (e.g. you would be able to sort the terms chronologically or by geographic region), and will give you far better control of your final output. I strongly suspect that unless you decided to self-publish, you would be asked by your prospective publisher to undertake this kind of exercise in any case, so you might as well organise your data in this way right from the start.
I suggest you lurk in some lexicographers' newsgroups for a while to find out exactly how the pros do it (they may contain archives that will point you in the right direction regarding data structure and software selection, or a relevant FAQ document); and if you can't find what you're after there, then you can always ask the other members directly. It is also probable that there are trade journals (either printed or online) which will answer your questions.
Reply from Erik Kowal ( - England)