The only instances I've seen of upper casing the first few words of a paragraph have been in the first paragraph of a chapter or scene - hence 'first paragraph'.
I doubt there's a regex engine that will differentiate between - 'LOS VEGAS IS CITY ONE SHOULD DODGE' and 'I HAD A GREAT TIME IN DODGE CITY.'
I just opened first three 'real' books in my Test library and looked at words 2-6 in the first paragraph of the first chapter. A factual history had no proper nouns, a Regency novel had 'HYDE PARK', and an international banking exposé had 'LONG-TERM CAPITAL MANAGEMENT'. Of those six words the only one that would even be flagged as a spelling error is 'hyde', the other five ('park', 'long', 'term', 'capital', and 'management') are valid in lower case - yet they would all be patently wrong in the context they are used.
Whilst ever there are no multi-lingual, deep-intelligence enabled regex engines which are capable of further deep-learning, us humans will always have something to do
BR