按空格和大写字母拆分文本 (PHP)

Question

I am trying to break the text by sentences. There are no dots in this text. But it contains capital letters. I use:

 <?php preg_match_all('/[A-Z][^A-Z]*?/Usu',$text,$sentences);

But it split the text only by capital letters. So I have such sentences as "S", "M", "S". It is wrong. I do not need to break such words as SMS. Help please.

Some clarification:

• I try to break the string before each string of one or more capital letters.
• But my real task is more complex. I am trying to format text for readability.
• Example: a piece of vacancy without html tags and line breaks: "Desirable: AWS experience Experience with Docker/Kubernetes". I try to get: "Desirable:", "AWS experience" and "Experience with Docker/Kubernetes" (I think I will be able to stick together very short strings after splitting by space and capital letter. Maybe it is a very bad way, of course).

Answer 1

You really shouldn't be using regex to parse something as complex as natural language. I'd recommend something like IntlBreakIterator instead.

$text = "Sentence 1. Sentence 2! Sentence 3? Sentence; number 4...Sentence, 5.";

$it = IntlBreakIterator::createSentenceInstance("en_US");
$it->setText($text);
$parts = $it->getPartsIterator();

foreach ($parts as $point => $sentence) {
    echo "$point => $sentence\n\n\n";
}

Output

0 => Sentence 1. 


1 => Sentence 2! 


2 => Sentence 3? 


3 => Sentence; number 4...


4 => Sentence, 5.

The rules for parsing words/sentences can be complex and daunting to implement in a regular expression. This solution is more sane for syntactically correct corpus. However, if the text has no punctuation like you say then there is no sane way to distinguish one sentence from another. Simply attempting to do it by capital letters can yield a lot of false positives because words can be capitalized mid-sentence such as proper nouns and some abbreviations.

Answer 2

I assume you you wish to break a string into pieces, where the break points are zero-width positions that immediately precede a capital letter and do not follow a capital letter. If so you could used the following regular expression.

(?=(?<![A-Z]|^)[A-Z])

Regex demo

The can be executed as follows:

<?php
$result = preg_split("/(?=(?<![A-Z]|^)[A-Z])/", "now is THE time to BE brave"); 
print_r($result); 

PHP demo

As shown at the link, this returns

Array
(
    [0] => now is 
    [1] => THE time to 
    [2] => BE brave
)

If the first word of the string were capitalized ( "Now" ), the first element of the string would be "Now is" (ie, not an empty string").

PHP's regex engine performs the following operations.

(?=           # begin a positive lookahead
  (?<!        # begin a negative lookbehind
    [A-Z]     # match a capital letter
    |         # or
    ^         # match the beginning of the line
  )           # end the negative lookbehind
  [A-Z]       # match a capital letter
)             # end positive lookahead

This attempts to match a capital letter in a positive lookahead ( [AZ] ), but that match fails if the negative lookbehind matches a capital letter preceding it or the capital letter is at the beginning of the string.