The only real constraints are sentences with first non-alpha. I'd try something like:
converted is store for modified text
new paragraph is store for modified sentences
abbreviations is list of common abbreviations (eg,nb,dr...)
paragraph is split by newline '\n'
for each newline in paragraph
/* abbreviation check using '.' seps */
if fullstops in newline greater than one
check newline for abbreviations and replace
/* break down para into newlines by '.' */
split newline into sentences by '.'
for sentence in sentences
if sentence starts with word
replace first letter with upper
else
move to first word, replace with upper
/* build up paragraphs */
append sentence to new paragraph
/* modified text */
append new paragraph to converted
write converted to display
tricky bit is 'abbreviations & dots' [0] & preserving the format. Will fail on names (eg: J.A. Smith)
