Create array of words from a string of text

Question

I would like to split a text into single words using PHP. Do you have any idea how to achieve this?

My approach:

function tokenizer($text) {
    $text = trim(strtolower($text));
    $punctuation = '/[^a-z0-9äöüß-]/';
    $result = preg_split($punctuation, $text, -1, PREG_SPLIT_NO_EMPTY);
    for ($i = 0; $i < count($result); $i++) {
        $result[$i] = trim($result[$i]);
    }
    return $result; // contains the single words
}
$text = 'This is an example text, it contains commas and full-stops. Exclamation marks, too! Question marks? All punctuation marks you know.';
print_r(tokenizer($text));

Is this a good approach? Do you have any idea for improvement?

Thanks in advance!

Fanny

Answer 1

Use the class \p{P} which matches any unicode punctuation character, combined with the \s whitespace class.

$result = preg_split('/((^\p{P}+)|(\p{P}*\s+\p{P}*)|(\p{P}+$))/', $text, -1, PREG_SPLIT_NO_EMPTY);

This will split on a group of one or more whitespace characters, but also suck in any surrounding punctuation characters. It also matches punctuation characters at the beginning or end of the string. This discriminates cases such as "don't" and "he said 'ouch!'"

Budde

+1, not sure, tho, how this will deal with äöüß. Does regex normally classify äöüß as word characters? – Carboy

Thank you. This would't probably work for English texts but I also want to extract German umlauts (ä, ö, ü), the "ß" and numbers in a string. The "\W" wouldn't extract "Fri3nd", would it? – Fanny

Seems it does not, but updated answer with something similar that works. – Budde

Updated answer works with perl (which php regex are based on): $ echo "äöüß, test" | perl -e 'while (<>) { if (/([\p{P}\s]+)/) { print "$1\n"; } }' , – Budde

Should one split don't into don and t? – Holleran

Updated it to handle such a case :) – Budde

Thanks, marcog, it works perfectly! But is it really better than my updated code above? Actually, what is the difference between our approaches? Is one faster than the other one? – Fanny

In your approach you're specifying the non-punctuation characters. You will be therefore be missing some cases, e.g. á. Why try manually specify them when the whole set of unicode punctuation characters has already been defined? And like eed3si9n pointed out with my original answer, yours will break up words such as don't. – Budde

@marcog Any idea what would be the Javascript equivalent of this? I tried doing str.split(/((^\p{P}+)|(\p{P}*\s+\p{P}*)|(\p{P}+$))/); but it doesn't work. Please help.. – Expellee

What to do if i want to remove all numbers? – Cowpuncher

Answer 2

Tokenize - strtok.

<?php
$text = 'This is an example text, it contains commas and full stops. Exclamation marks, too! Question marks? All punctuation marks you know.';
$delim = ' \n\t,.!?:;';

$tok = strtok($text, $delim);

while ($tok !== false) {
    echo "Word=$tok<br />";
    $tok = strtok($delim);
}
?>

Holleran

This won't work if you get a : or ; or any other punctuation character you haven't accounted for. – Budde

@marcog, I added : and ;. Doesn't {P} catch apostrophe and hyphen? – Holleran

What about cases such quoting? My updated answer discriminates between these cases. – Budde

Excellent idea. Added +1. The only thing is that there should be double quotes around $delim = " \n\t,.!?:;"; With the single quotes it does not work correctly, it splits by the letter n too. – Ledet

Answer 3

I would first make the string to lower-case before splitting it up. That would make the i modifier and the array processing afterwards unnecessary. Additionally I would use the \W shorthand for non-word characters and add a + multiplier.

$text = 'This is an example text, it contains commas and full stops. Exclamation marks, too! Question marks? All punctuation marks you know.';
$result = preg_split('/\W+/', strtolower($text), -1, PREG_SPLIT_NO_EMPTY);

---

Edit   Use the Unicode character properties instead of \W as marcog suggested. Something like [\p{P}\p{Z}] (punctuation and separator characters) would cover the characters more specific than \W.

Ernaldus

Thanks, the idea to perform strtolower() before is very good. I'll use this. – Fanny

What purpose does strtolower() serve if you are splitting with \W? Do you want to add a u pattern modifier? A note to researchers... \W will not match an underscore. – Modify

Answer 4

Do:

str_word_count($text, 1);

Or if you need unicode support:

function str_word_count_Helper($string, $format = 0, $search = null)
{
    $result = array();
    $matches = array();

    if (preg_match_all('~[\p{L}\p{Mn}\p{Pd}\'\x{2019}' . preg_quote($search, '~') . ']+~u', $string, $matches) > 0)
    {
        $result = $matches[0];
    }

    if ($format == 0)
    {
        return count($result);
    }

    return $result;
}

Oleviaolfaction

Thanks but this wouldn't work. "Fri3nd" wouldn't be extracted but it should. – Fanny

I don't understand why "Fri3nd" should be extracted. Removed from the array, broken down into "Fri3" and "nd" (or similar)? O.o – Chokefull

If you want to consider numbers as words just do str_word_count_Helper($string, 1, '0123456789'); – Oleviaolfaction

Native PHP functions that allow double-dot range syntax demonstrates that str_word_count($string, 1, '0..9') will do. – Modify

Answer 5

you can also use PHP strtok() function to fetch string tokens from your large string. you can use it like this:

 $result = array();
 // your original string
 $text = 'This is an example text, it contains commas and full stops. Exclamation marks, too! Question marks? All punctuation marks you know.';
 // you pass strtok() your string, and a delimiter to specify how tokens are separated. words are seperated by a space.
 $word = strtok($text,' ');
 while ( $word !== false ) {
     $result[] = $word;
     $word = strtok(' ');
 }

see more on php documentation for strtok()

Polled

what is the difference between this and explode(' ', $text); – Vacation

The code sample in the question is a tokenizer, my answer was implying that PHP has a string tokenizer built-in. Also explode() will return all of the words of the text at once, but using strtok() the caller has the choice to stop searching for words in the text, as soon as a desired condition is met. Other than this, I can't think of any other difference. – Polled

Answer 6

You can also use the method explode : http://php.net/manual/en/function.explode.php

$words = explode(" ", $sentence);

Laubin

not works with 2 or more consecutive spaces. you have to use a foreach with explode(" ", $sentence) within if($word == "") continue; so you could avoid empty words. – Bedpost