在英语中，有乐趣和之间的简单区别an和a：使用an开始以元音字前面时，和a当单词以辅音开始。

为了简单起见，an在以元音（aeiou）a开头的单词之前，以辅音开头的单词之前。

输入值

仅包含可打印ASCII字符的字符串，[?]出现在必须选择插入an或的位置a。[?]总是出现在一个字之前。您可以假定该句子在语法上是正确的，并且格式与通常一样。

输出量

[?]用适当的单词（an或a）替换的输入字符串。您不必担心大写！

何时大写

如果单词前面没有字符（输入中的第一个），或者.?!后面带有空格，则将其大写。

例子

Input: Hello, this is [?] world!
Output: Hello, this is a world!

Input: How about we build [?] big building. It will have [?] orange banana hanging out of [?] window.
Output: How about we build a big building. It will have an orange banana hanging out of a window.

Input: [?] giant en le sky.
Output: A giant en le sky.

Input: [?] yarn ball? [?] big one!
Output: A yarn ball? A big one!

Input: [?] hour ago I met [?] European.
Output: A hour ago I met an European.

Input: Hey sir [Richard], how 'bout [?] cat?
Output: Hey sir [Richard], how 'bout a cat?

这是代码高尔夫球，因此以字节为单位的最短代码胜出！

V，41字节

ÍãÛ?Ý ¨[aeiou]©/an
ÍÛ?Ý/a
Í^aü[.!?] a/A

在线尝试！，也可以方便地用于验证所有测试用例，而无需额外的字节数。

这利用了V的“正则表达式压缩”。它使用了许多不可打印的字符，因此这是一个十六进制转储：

0000000: cde3 db3f dd85 20a8 5b61 6569 6f75 5da9  ...?.. .[aeiou].
0000010: 2f61 6e0a cddb 3fdd 2f61 0acd 5e61 fc5b  /an...?./a..^a.[
0000020: 2e21 3f5d 2093 612f 41                   .!?] .a/A

Perl，48个字节

由于Ton Hospel节省了1个字节。

#!perl -p
s;\[\?];A.n x$'=~/^ [aeiou]/i^$"x/[^.?!] \G/;eg

将shebang视为一个，输入来自stdin。

说明

#!perl -p               # for each line of input, set $_, auto-print result

s;                      # begin regex substitution, with delimiter ;
\[\?]                   # match [?] literally, and replace with:
;
A.n x$'=~/^ [aeiou]/i   # 'A', concatenate with 'n' if post-match ($')
                        #   matches space followed by a vowel
^$"x/[^.?!] \G/         # if the match is preceded by /[^.?!] /, xor with a space
                        #   this will change An -> an

;eg                     # regex options eval, global

样品用量

$ echo Hello, this is [?] world! | perl a-an.pl
Hello, this is a world!

$ echo How about we build [?] big building. It will have [?] orange banana hanging out of [?] window. | perl a-an.pl
How about we build a big building. It will have an orange banana hanging out of a window.

$ echo [?] giant en le sky. [?] yarn ball? | perl a-an.pl
A giant en le sky. A yarn ball?

$ echo [?] hour ago I met [?] European. | perl a-an.pl
A hour ago I met an European.

Ruby，78 72字节

->s{s.gsub(/(^|\. )?\K\[\?\]( [aeiou])?/i){"anAn"[$1?2:0,$2?2:1]+"#$2"}}

• @Jordan节省了6个字节

不打高尔夫球

def f(s)
    s.gsub(/(^|\. )?\[\?\]( [aeiou])?/i) do |m|
        capitalize = $1
        vowel = $2
        replacement = if vowel then
            capitalize ? "An" : "an"
        else
            capitalize ? "A" : "a"
        end
        m.sub('[?]', replacement)
    end
end

PHP，207字节

foreach(explode("[?]",$s)as$i=>$b){$r=Aa[$k=0|!strstr(".!?",''==($c=trim($a))?".":$c[strlen($c)-1])].n[!preg_match("#^['\"´`\s]*([aeiou]|$)#i",$d=trim($b))];echo$i?$r.$b:$b;$a=$i?''==$d?a:$b:(''==$d?".":a);}

我喜欢不时提供更完整的解决方案...
但是我必须承认，尽管这还没有，但是这有点过头了。

保存到文件，使用php <filename>STDIN的输入运行。

测试用例

How about we build [?] big building ... with [?] orange banana hanging out of [?] window.
=>  How about we build a big building ... with an orange banana hanging out of a window.

Hello, this is [?] world!               =>  Hello, this is a world!
Should I use [?] '[?]' or [?] '[?]'?    =>  Should I use an 'an' or an 'a'?
[?] elephant in [?] swimsuit.           =>  An elephant in a swimsuit.

How I met your moth[?].                 =>  How I met your motha.
b[?][?][?] short[?]ge!                  =>  banana shortage!

分解

foreach(explode("[?]",$s)as$i=>$b)
{
    $r=
        // lookbehind: uppercase if the end of a sentence precedes
        Aa[$k=0|!strstr(".!?",''==($c=trim($a))?".":$c[strlen($c)-1])]
        .
        // lookahead: append "n" if a vowel follows (consider quote characters blank)
        n[!preg_match("#^['\"´`\s]*([aeiou]|$)#i",$d=trim($b))]
    ;
    // output replacement and this part
    echo$i?$r.$b:$b;
    // prepare previous part for next iteration
    $a=$i               // this part was NOT the first:
        ?   ''==$d
            ? a             // if empty -> a word ($r from the previous iteration)
            : $b            // default: $b
        :  (''==$d      // this WAS the first part:
            ? "."           // if empty: end of a sentence (= uppercase next $r)
            : a             // else not
        )
    ;
    // golfed down to `$a=!$i^''==$d?a:($i?$b:".");`
}

Minkolang 0.15、75字节

od4&r$O."]?["30$Z3&00w4X"Aa"I2-"Aa ."40$Z,*2&$rxr$O" aeiou"od0Z1=3&"n"r5X$r

在这里尝试！

说明

od                                                                    Take character from input and duplicate (0 if input is empty)
  4&                                                                  Pop top of stack; jump 4 spaces if not 0
    r$O.                                                              Reverse stack, output whole stack as characters, and stop.

    "]?["                                                             Push "[?]" on the stack
         30$Z                                                         Pop the top 3 items and count its occurrences in the stack
              3&                                                      Pop top of stack; jump 3 spaces if not 0
                00w                                                   Wormhole to (0,0) in the code box

                3X                                                    Dump the top 3 items of stack
                  "Aa"                                                Push "aA"
                      I2-                                             Push the length of stack minus 2
                         "Aa ."40$Z,                                  Push ". aA" and count its occurrences, negating the result
                                    *                                 Multiply the top two items of the stack
                                     2&$r                             Pop top of stack and swap the top two items if 0
                                         x                            Dump top of stack
                                          r                           Reverse stack
                                           $O                         Output whole stack as characters
                                             " aeiou"                 Push a space and the vowels
                                                     od               Take a character from input and duplicate
                                                       0Z             Pop top of stack and count its occurrences in the stack (either 1 or 2)
                                                         1=           1 if equal to 1, 0 otherwise
                                                           3&         Pop top of stack; jump 3 spaces if not 0
                                                             "n"      Push "n" if top of stack is 0

                                                             r        Reverse stack
                                                              5X      Dump top five items of stack
                                                                $r    Swap top two items of stack

请注意，由于Minkolang是环形的，因此当程序计数器移出右侧边缘时，它将再次出现在左侧。当然可以打高尔夫球，但是由于规格原因我不得不添加21个字节，因此我可能不会尝试。

JavaScript（ES6），90 86 87 85

更改大写规范后再次编辑（现在更明智）

再次编辑 1个字节，保存@Huntro

编辑正如IsmaelMiguel指出的那样，再 2个字节以管理引号等（即使我不知道op是否要求这样做）。请注意，以前我已经计算了86个字节，但它们是85个

如果不完整（至少），则尝试遵循注释事件中规定的大写规则

x=>x.replace(/([^!?.] )?\[\?](\W*.)/g,(a,b,c)=>(b?b+'a':'A')+(/[aeiou]/i.test(c)?'n'+c:c))

测试

f=x=>x.replace(/([^!?.] )?\[\?](\W*.)/g,(a,b,c)=>(b?b+'a':'A')+(/[aeiou]/i.test(c)?'n'+c:c))

function go() {
  var i=I.value, o=f(i)
  O.innerHTML = '<i>'+i+'</i>\n<b>'+o+'</b>\n\n'+O.innerHTML 
}

go()

#I { width:80% }

<input value='How about we build [?] big building. It will have [?] orange banana hanging out of [?] window.' id=I><button onclick='go()'>GO</button><pre id=O></pre>

批处理，136字节

@set/ps=
@for %%v in (a e i o u)do @call set s=%%s:[?] %%v=an %%v%%
@set s=%s:[?]=a%
@if %s:~0,1%==a set s=A%s:~1%
@echo %s:. a=. A%

在STDIN上输入一行。

PHP, 100 92 bytes

<?=preg_filter(["/\[\?]\K(?= [aeiou])/i","/([.?!] |^)\K\[\?]/","/\[\?]/"],[n,A,a],$argv[1]);

It was possible to further golf the regular expressions.

Gives a notice about an undefined constant but still works.

Edit: 8 bytes saved thanks to primo

Python 3.5.1, 153 147 124 Bytes

*s,=input().replace('[?]','*');print(*[('a','A')[i<1or s[i-2]in'.?!']+'n'*(s[i+2]in 'aeiouAEIOU')if c=='*'else c for i,c in enumerate(s)],sep='')

Input :

[?] apple [?] day keeps the doctor away. [?] lie.

Output :

An apple a day keeps the doctor away. A lie.

123 Bytes version - This does not handle capitalization rule.

s=list(input().replace('[?]','*'));print(*['a'+'n'*(s[i+2]in 'aeiouAEIOU')if c=='*'else c for i,c in enumerate(s)],sep='')

Ideone it!

Java, 180178 bytes

My first post here, I did use a part of the Kevin Cruijssen post but and up with a different approach, he helped me to reduce a bit more so, thanks to him !

String c(String s){String x[]=s.split("\\[\\?]",2),r=x[0];return x.length>1?r+(r.matches("(.+[.!?] )|(^)$")?"A":"a")+("aeiouAEIOU".contains(""+x[1].charAt(1))?"n":"")+c(x[1]):r;}

Here it is ungolfed :

static String c(String s) {
        String x[] = s.split("\\[\\?\\]", 2), r = x[0];
        return x.length > 1 ? r + (r.matches("(.+[.!?] )|(^)$") ? "A" : "a")
                + ("aeiouAEIOU".contains("" + x[1].charAt(1)) ? "n" : "") + c(x[1]) : r;
    }

And the result

A simple explanation, I use a recursive approch to find every [?].

I couldn't find a way to use the matches with insensitive case (not sure it is possible).

178bytes : Thanks to Martin Ender !

05AB1E, 38 36 35 bytes

2FžNžM‚NèSðì…[?]©ìDu«D®'a'nN׫::}.ª

Try it online or verify all test cases.

Explanation:

2F            # Loop 2 times:
  žN          #  Push consonants "bcdfghjklmnpqrstvwxyz"
  žM          #  Push vowels "aeiou"
    ‚         #  Pair them together into a list
     Nè       #  And use the loop-index to index into this pair
  S           #  Convert this string to a list of characters
   ðì         #  Prepend a space in front of each character
     …[?]     #  Push string "[?]
         ©    #  Store it in variable `®` (without popping)
          ì   #  And prepend it in front of each string in the list as well
  }D          #  Then duplicate the list
    u         #  Uppercase the characters in the copy
     «        #  And merge the two lists together
              #   i.e. for the vowel-iteration we'd have ["[?] a","[?] e","[?] i","[?] o",
              #    "[?] u","[?] A","[?] E","[?] I","[?] O","[?] U"]
   D          #  Duplicate it
    ®         #  Push "[?]" from variable `®`
     'a      '#  Push "a"
       'n    '#  Push "n"
         N×   #  Repeated the 0-based index amount of times (so either "" or "n")
           «  #  And append it to the "a"
    :         #  Replace all "[?]" with "an"/"a" in the duplicated list
     :        #  And then replace all values of the lists in the (implicit) input-string
 }.ª          #  After the loop: sentence-capitalize everything (which fortunately retains
              #  capitalized words in the middle of sentences, like the "European" testcase)
              # (and after the loop the result is output implicitly)

C#, 204 235 bytes

string n(string b){for(int i=0;i<b.Length;i++){if(b[i]=='['){var r="a";r=i==0||b[i-2]=='.'?"A":r;r=System.Text.RegularExpressions.Regex.IsMatch(b[i+4].ToString(),@"[aeiouAEIOU]")?r+"n":r;b=b.Insert(i+3,r);}}return b.Replace("[?]","");}

Ungolfed full program:

using System;

class a
{
    static void Main()
    {
        string s = Console.ReadLine();
        a c = new a();
        Console.WriteLine(c.n(s));
    }

    string n(string b)
    {
        for (int i = 0; i < b.Length; i++)
        {
            if (b[i] == '[')
            {
                var r = "a";
                r = i == 0 || b[i - 2] == '.' ? "A" : r;
                r = System.Text.RegularExpressions.Regex.IsMatch(b[i + 4].ToString(), @"[aeiouAEIOU]") ? r + "n" : r;
                b = b.Insert(i + 3, r);
            }
        }
        return b.Replace("[?]", "");
    }
}

I'm sure this could be improved, especially the Regex part, but can't think of anything right now.

Java 7, 239 214 213 bytes

String c(String s){String x[]=s.split("\\[\\?\\]"),r="";int i=0,l=x.length-1;for(;i<l;r+=x[i]+(x[i].length()<1|x[i].matches(".+[.!?] $")?65:'a')+("aeiouAEIOU".contains(x[++i].charAt(1)+"")?"n":""));return r+x[l];}

Ungolfed & test cases:

Try it here.

class M{
  static String c(String s){
    String x[] = s.split("\\[\\?\\]"),
           r = "";
    int i = 0,
        l = x.length - 1;
    for (; i < l; r += x[i]
                     + (x[i].length() < 1 | x[i].matches(".+[.!?] $") 
                        ? 65
                        : 'a')
                     + ("aeiouAEIOU".contains(x[++i].charAt(1)+"")
                        ? "n"
                        : ""));
    return r + x[l];
  }

  public static void main(String[] a){
    System.out.println(c("Hello, this is [?] world!"));
    System.out.println(c("How about we build [?] big building. It will have [?] orange banana hanging out of [?] window."));
    System.out.println(c("[?] giant en le sky."));
    System.out.println(c("[?] yarn ball? [?] big one!"));
    System.out.println(c("[?] hour ago I met [?] European. "));
    System.out.println(c("Hey sir [Richard], how 'bout [?] cat?"));
    System.out.println(c("[?] dog is barking. [?] cat is scared!"));
  }
}

Output:

Hello, this is a world!
How about we build a big building. It will have an orange banana hanging out of a window.
A giant en le sky.
A yarn ball? A big one!
A hour ago I met an European. 
Hey sir [Richard], how 'bout a cat?
A dog is barking. A cat is scared!

Racket 451 bytes (without regex)

It is obviously a long answer but it replaces a and an with capitalization also:

(define(lc sl item)(ormap(lambda(x)(equal? item x))sl))
(define(lr l i)(list-ref l i))(define(f str)(define sl(string-split str))
(for((i(length sl))#:when(equal?(lr sl i)"[?]"))(define o(if(lc(string->list"aeiouAEIOU")
(string-ref(lr sl(add1 i))0))#t #f))(define p(if(or(= i 0)(lc(string->list".!?")
(let((pr(lr sl(sub1 i))))(string-ref pr(sub1(string-length pr))))))#t #f))
(set! sl(list-set sl i(if o(if p"An""an")(if p"A""a")))))(string-join sl))

Testing:

(f "[?] giant en le [?] sky.")
(f "[?] yarn ball?")
(f "[?] hour ago I met [?] European. ")
(f "How about we build [?] big building. It will have [?] orange banana hanging out of [?] window.")
(f "Hello, this is [?] world!")

Output:

"A giant en le a sky."
"A yarn ball?"
"A hour ago I met an European."
"How about we build a big building. It will have an orange banana hanging out of a window."
"Hello, this is a world!"

Detailed version:

(define(contains sl item)
  (ormap(lambda(x)(equal? item x))sl))

(define(lr l i)
  (list-ref l i))

(define(f str)
  (define sl(string-split str))
  (for((i(length sl))#:when(equal?(lr sl i)"[?]"))
    (define an   ; a or an
      (if(contains(string->list "aeiouAEIOU")
                  (string-ref(lr sl(add1 i))0))
         #t #f ))
    (define cap   ; capital or not
      (if(or(= i 0)(contains(string->list ".!?")
                            (let ((prev (lr sl(sub1 i)))) (string-ref prev
                                       (sub1(string-length prev))))))
         #t #f))
    (set! sl(list-set sl i (if an (if cap "An" "an" )
                                 (if cap "A" "a")))))
  (string-join sl))

J, 113 bytes

[:;:inv 3(0 2&{(((('aA'{~[)<@,'n'#~])~('.?!'e.~{:))~('AEIOUaeiou'e.~{.))&>/@[^:(<@'[?]'=])1{])\' 'cut' . '([,~,)]

Try it online!

Shame, shame!

Retina, 66 60 bytes

i`\[\?\]( ([aeiou]?)[a-z&&[^aeiou])
a$.2*n$1
(^|[.?!] )a
$1A

Try it online.

Explanation:

Do a case-insensitive search for [?] followed by a vowel or consonant, where the optional vowel is saved in capture group 2, and the entire match in capture group 1:

i`\[\?\]( ([aeiou]?)[a-z&&[^aeiou])

Replace this with an a, followed by the length of the second group amount of n (so either 0 or 1 n), followed by the letter(s) of capture group 1:

a$.2*n$1

Then match an a at either the start of the string, or after either of .?! plus a space:

(^|[.?!] )a

And uppercase that A, without removing the other characters of capture group 1:

$1A

Java (JDK), 154 bytes

s->{String v="(?= [aeiou])",q="(?i)\\[\\?]",b="(?<=^|[?.!] )";return s.replaceAll(b+q+v,"An").replaceAll(q+v,"an").replaceAll(b+q,"A").replaceAll(q,"a");}

Try it online!

Explanation:

s->{
    String v="(?= [aeiou])",          // matches being followed by a vowel
    q="(?i)\\[\\?]",                  // matches being a [?]
    b="(?<=^|[?.!] )";                // matches being preceded by a sentence beginning
    return s.replaceAll(b+q+v,"An")   // if beginning [?] vowel, you need "An"
        .replaceAll(q+v,"an")         // if           [?] vowel, you need "an"
        .replaceAll(b+q,"A")          // if beginning [?]      , you need "A"
        .replaceAll(q,"a");}          // if           [?]      , you need "a"

C (gcc), 225 207 202 201 bytes

Thanks to ceilingcat for -24 bytes

#define P strcpy(f+d,index("!?.",i[c-2])+!c?
c;d;v(i,g,f)char*i,*g,*f;{for(d=0;i[c];c++,d++)strcmp("[?]",memcpy(g,i+c,3))?f[d]=i[c]:(index("aeiouAEIOU",i[c+4])?P"An ":"an "),d++:P"A ":"a "),d++,c+=3);}

Try it online!

Groovy, 73 162 bytes

def a(s){s.replaceAll(/(?i)(?:(.)?( )?)\[\?\] (.)/){r->"${r[1]?:''}${r[2]?:''}${'.?!'.contains(r[1]?:'.')?'A':'a'}${'aAeEiIoOuU'.contains(r[3])?'n':''} ${r[3]}"}}

edit: damn, the capitalization totally complicated everything here

C# 209 bytes

string A(string b){var s=b.Split(new[]{"[?]"},0);return s.Skip(1).Aggregate(s[0],(x,y)=>x+(x==""||(x.Last()==' '&&".?!".Contains(x.Trim().Last()))?"A":"a")+("AEIOUaeiou".Contains(y.Trim().First())?"n":"")+y);}

Formatted

string A(string b)
{
    var s = b.Split(new[] { "[?]" }, 0);
    return s.Skip(1).Aggregate(s[0], (x, y) => x + (x == "" || (x.Last() == ' ' && ".?!".Contains(x.Trim().Last())) ? "A" : "a") + ("AEIOUaeiou".Contains(y.Trim().First()) ? "n" : "") + y);
}

Perl 6, 78 bytes

{S:i:g/(^|<[.?!]>' ')?'[?] '(<[aeiou]>?)/{$0 xx?$0}{<a A>[?$0]}{'n'x?~$1} $1/}

Explanation:

{
  S
    :ignorecase
    :global
  /
    ( # $0
    | ^             # beginning of line
    | <[.?!]> ' '   # or one of [.?!] followed by a space
    ) ?             # optionally ( $0 will be Nil if it doesn't match )

    '[?] '          # the thing to replace ( with trailing space )

    ( # $1
      <[aeiou]> ?   # optional vowel ( $1 will be '' if it doesn't match )
    )

  /{
    $0 xx ?$0      # list repeat $0 if $0
                   # ( so that it doesn't produce an error )
  }{
    < a A >[ ?$0 ] # 'A' if $0 exists, otherwise 'a'
  }{
    'n' x ?~$1     # 'n' if $1 isn't empty
                   # ｢~｣ turns the Match into a Str
                   # ｢?｣ turns that Str into a Bool
                   # ｢x｣ string repeat the left side by the amount of the right

  # a space and the vowel we may have borrowed
  } $1/
}

Test:

#! /usr/bin/env perl6
use v6.c;
use Test;

my &code = {S:i:g/(^|<[.?!]>' ')?'[?] '(<[aeiou]>?)/{<a A>[?$0]~('n'x?~$1)} $1/}

my @tests = (
  'Hello, this is [?] world!'
  => 'Hello, this is a world!',

  'How about we build [?] big building. It will have [?] orange banana hanging out of [?] window.'
  => 'How about we build a big building. It will have an orange banana hanging out of a window.',

  '[?] giant en le sky.'
  => 'A giant en le sky.',

  '[?] yarn ball?'
  => 'A yarn ball?',

  '[?] hour ago I met [?] European.'
  => 'A hour ago I met an European.',

  "Hey sir [Richard], how 'bout [?] cat?"
  => "Hey sir [Richard], how 'bout a cat?",
);

plan +@tests;

for @tests -> $_ ( :key($input), :value($expected) ) {
  is code($input), $expected, $input.perl;
}

1..6
ok 1 - "Hello, this is a world!"
ok 2 - "How about we build a big building. It will have an orange banana hanging out of a window."
ok 3 - "A giant en le sky."
ok 4 - "A yarn ball?"
ok 5 - "A hour ago I met an European."
ok 6 - "Hey sir [Richard], how 'bout a cat?"

Lua, 131 Bytes.

function(s)return s:gsub("%[%?%](%s*.)",function(a)return"a"..(a:find("[AEIOUaeiou]")and"n"or"")..a end):gsub("^.",string.upper)end

Although lua is a terrible Language for golfing, I feel I've done pretty well.

Pip, 62 55 54 50 bytes

Takes the string as a command-line argument.

aR-`([^.?!] )?\[\?]( [^aeiou])?`{[b"aA"@!b'nX!cc]}

Try it online!

Explanation:

a                                                   Cmdline argument
 R                                                  Replace...
  -`                           `                    The following regex (case-insensitive):
    ([^.?!] )?                                      Group 1: not end-of-sentence (nil if it doesn't match)
              \[\?]                                 [?]
                   ( [^aeiou])?                     Group 2: not vowel (nil if there is a vowel)
                                {                }  ... with this callback function (b = grp1, c = grp2):
                                 [              ]   List (concatenated when cast to string) of:
                                  b                 Group 1
                                   "aA"@!b          "a" if group 1 matched, else "A"
                                          'nX!c     "n" if group 2 didn't match, else ""
                                               c    Group 2

Racket (with regex) 228 bytes

(define(r a b c)(regexp-replace* a b c))
(define(f s)
(set! s(r #rx"[a-zA-Z ]\\[\\?\\] (?=[aeiouAEIOU])"s" an "))
(set! s(r #rx"[a-zA-Z ]\\[\\?\\]"s" a"))
(set! s(r #rx"\\[\\?\\] (?=[aeiouAEIOU])"s"An "))
(r #rx"\\[\\?\\]"s"A"))

Testing:

(f "[?] giant en le [?] sky.")
(f "[?] yarn ball?")
(f "[?] apple?")
(f "[?] hour ago I met [?] European. ")
(f "How about we build [?] big building. It will have [?] orange banana hanging out of [?] window.")
(f "Hello, this is [?] world!")

Output:

"A giant en le a sky."
"A yarn ball?"
"An apple?"
"A hour ago I met an European. "
"How about we build a big building. It will have an orange banana hanging out of a window."
"Hello, this is a world!"

Python 3, 104 103 bytes

-1 bytes, unescaped ]

lambda s:r('(^|[.?!] )a',r'\1A',r('a( [aeiouAEIOU])',r'an\1',r('\[\?]','a',s)));from re import sub as r

Try it online!

Starts by replacing all occurences of [?] with a,
Then replaces all a followed by a vowel, with an.
Then replaces all a at the start of input or a sentence with A.

Assumes that [?] will never be touching another word, and that lower-case a should never begin a sentence.

PowerShell, 124 bytes

inspired by Avi's answer for Java.

$args-replace(($b='(?<=^|[?.!] )')+($q='\[\?]')+($v='(?= [aeiou])')),'An'-replace"$q$v",'an'-replace"$b$q",'A'-replace$q,'a'

Try it online!