Extract Keywords From A Text String With PHP

Wednesday, July 29, 2009 - 09:27

A common issue I have come across in the past is that I have a CMS system, or an old copy of Wordpress, and I need to create a set of keywords to be used in the meta keywords field. To solve this I put together a simple function that runs through a string and picks out the most commonly used words in that list as an array. This is currently set to be 10, but you can change that quite easily.

The first thing the function defines is a list of "stop" words. This is a list of words that occur quite a bit in English text and would therefore interfere with the outcome of the function. The function also uses a variant of the slug function to remove any odd characters that might be in the text.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
function extractCommonWords($string){
      $stopWords = array('i','a','about','an','and','are','as','at','be','by','com','de','en','for','from','how','in','is','it','la','of','on','or','that','the','this','to','was','what','when','where','who','will','with','und','the','www');
   
      $string = preg_replace('/\s\s+/i', '', $string); // replace whitespace
      $string = trim($string); // trim the string
      $string = preg_replace('/[^a-zA-Z0-9 -]/', '', $string); // only take alphanumerical characters, but keep the spaces and dashes too…
      $string = strtolower($string); // make it lowercase
   
      preg_match_all('/\b.*?\b/i', $string, $matchWords);
      $matchWords = $matchWords[0];
      
      foreach ( $matchWords as $key=>$item ) {
          if ( $item == '' || in_array(strtolower($item), $stopWords) || strlen($item) <= 3 ) {
              unset($matchWords[$key]);
          }
      }   
      $wordCountArr = array();
      if ( is_array($matchWords) ) {
          foreach ( $matchWords as $key => $val ) {
              $val = strtolower($val);
              if ( isset($wordCountArr[$val]) ) {
                  $wordCountArr[$val]++;
              } else {
                  $wordCountArr[$val] = 1;
              }
          }
      }
      arsort($wordCountArr);
      $wordCountArr = array_slice($wordCountArr, 0, 10);
      return $wordCountArr;
}

The function returns the 10 most commonly occurring words as an array, with the key as the word and the amount of times it occurs as the value. To extract the words just use the implode() function in conjunction with the array_keys() function. To change the number of words returned just alter the value in the third parameter of the array_slice() function near the return statement, currently set to 10. Here is an example of the function in action.

1
2
3
$text = "This is some text. This is some text. Vending Machines are great.";
$words = extractCommonWords($text);
echo implode(',', array_keys($words));

This produces the following output.

some,text,machines,vending
Category: 
philipnorton42's picture

Philip Norton

Phil is the founder and administrator of #! code and is an IT professional working in the North West of the UK.
Google+ | Twitter

Comments

hello can any one help me with a problem? I have a variable containing a html commando like this: $str="<a id=home href=http://d-grund.dk/ title=D-Grund.DK target=_blank><img id=home src=http://d_grund.dk/images/homelogo.gif alt=D-Grund.DK border=1 width=166 height=90 /></a>"; and i will split it up so i can change the values and sample it again to work with echo $str; sorry for my bad english. i have searched the entire net and nothing is what i need so i hope anyone can help me. my solution is not good because it bet big and time consuming. and i have find some on the internet that look a like that i need but i cant sample it again without destroying anything. sincerely LAT D-Grund.dk Administrator my email is lat@ofir.dk
philipnorton42's picture
Submitted by philipnorton42 on Sun, 06/20/2010 - 18:28

I would love to help but I'm not sure what you are trying to do. Passing this string through the extractCommonWords() function probably won't produce any meaningful output as the string is just HTML containing a link and an image.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
function extractCommonWords($string){
      $stopWords = array('i','a','about','an','and','are','as','at','be','by','com','de','en','for','from','how','in','is','it','la','of','on','or','that','the','this','to','was','what','when','where','who','will','with','und','the','www');
   
      $string = preg_replace('/ss+/i', '', $string);
      $string = trim($string); // trim the string
      $string = preg_replace('/[^a-zA-Z0-9 -]/', '', $string); // only take alphanumerical characters, but keep the spaces and dashes too…
      $string = strtolower($string); // make it lowercase
   
      preg_match_all('/\b.*?\b/i', $string, $matchWords);
      $matchWords = $matchWords[0];
      
      foreach ( $matchWords as $key=>$item ) {
          if ( $item == '' || in_array(strtolower($item), $stopWords) || strlen($item) <= 3 ) {
              unset($matchWords[$key]);
          }
      }   
      $wordCountArr = array();
      if ( is_array($matchWords) ) {
          foreach ( $matchWords as $key => $val ) {
              $val = strtolower($val);
              if ( isset($wordCountArr[$val]) ) {
                  $wordCountArr[$val]++;
              } else {
                  $wordCountArr[$val] = 1;
              }
          }
      }
      arsort($wordCountArr);
      $wordCountArr = array_slice($wordCountArr, 0, 10);
      return $wordCountArr;
  }
}
 
$text = "This is some text. This is some text. Vending Machines are great.";
$words = extractCommonWords($text);
echo implode(',', array_keys($words));
philipnorton42's picture
Submitted by philipnorton42 on Fri, 07/02/2010 - 15:15

Thanks for the input, after running the script it was clear that it wouldn't work as there was one too many curly braces in the function. I have removed this in the example so it will run now.
Do you maybe know, what would be the problem, .. when I get my keywords with your script.. I always get letter "p" added right before first keyword and right after last keyword.. os the keyword list looks like this: pkeyword1, keyword2...... keyword10p .. and I dont know what causes this.. thanks
philipnorton42's picture
Submitted by philipnorton42 on Tue, 07/20/2010 - 09:51

Are you passing it a HTML string with p tags at the start and end? Try using strip_tags() first.
Your approach of stripping punctuation was a heck of a lot better than my method. I went the opposite approach and used a stock snippet of code with several preg_replace patterns, except for WordPress curly smart quotes. This is much more elegant and simple. Thanks for that, and here is how I would count word frequency:
1
    $wordCountArray = array_count_values( $matchWords );
No need of a foreach loop. The strtolower() call inside both foreach loops is really not needed either since you have already converted $string to lowercase beforehand.
Hey I’ve found your code very useful. Thanks a lot! However I’m finding an issue and that is that I want to divide the number of times a word repeats by the total number of words. The problem I have is that I cannot access each item of the array since the keys change every time I change the text. I was wondering if there is a say around this or if a new dimension can be added to the array so that things like this will work: echo $wordCountArr[2]/$totalwords; Any help will be much appreciated! Thanks in advance! :)
By the way, I've also tryed this with now luck... $numrepeats = print_r (array_values($words), true); for ($h=0; $h"; }
philipnorton42's picture
Submitted by philipnorton42 on Wed, 01/19/2011 - 11:18

@Guillermo I think to get the effect you are looking for you'll need to modify the function in the following way:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
function extractCommonWords($string){
      $stopWords = array('i','a','about','an','and','are','as','at','be','by','com','de','en','for','from','how','in','is','it','la','of','on','or','that','the','this','to','was','what','when','where','who','will','with','und','the','www');
   
      $string = preg_replace('/ss+/i', '', $string);
      $string = trim($string); // trim the string
      $string = preg_replace('/[^a-zA-Z0-9 -]/', '', $string); // only take alphanumerical characters, but keep the spaces and dashes too…
      $string = strtolower($string); // make it lowercase
   
      preg_match_all('/\b.*?\b/i', $string, $matchWords);
      $matchWords = $matchWords[0];
 
      $totalWords = count($matchWords[0]);
 
      foreach ( $matchWords as $key=>$item ) {
          if ( $item == '' || in_array(strtolower($item), $stopWords) || strlen($item) <= 3 ) {
              unset($matchWords[$key]);
          }
      }
      $wordCountArr = array();
      if ( is_array($matchWords) ) {
          foreach ( $matchWords as $key => $val ) {
              $val = strtolower($val);
              if ( !isset($wordCountArr[$val]) {
                  $wordCountArr[$val] = array();
              }
              if ( isset($wordCountArr[$val]['count']) ) {
                  $wordCountArr[$val]['count']++;
              } else {
                  $wordCountArr[$val]['count'] = 1;
              }
          }
          arsort($wordCountArr);
          $wordCountArr = array_slice($wordCountArr, 0, 10);
          foreach ( $wordCountArr as $key => $val) {
              $wordCountArr[$val]['bytotal'] = $wordCountArr[$val]['count'] / $totalWords;
          } 
      }
      return $wordCountArr;
}
This is untested code, but it should get you what you need. :)
Hey thanks for your super-fast response! I get it quite to work (not saying the code is not right, I just didn’t figure out how to implement it). But I’ve found a way around the issue. Thank you very much!
Excellent bit of code. I was looking for something to auto-suggest keywords for a content management system. I made the following changes : 1) removed the strtolower call in the foreach loop because the string has already been forced to lower case so all the items will be lower case anyway.
1
if ( $item == '' || in_array($item, $stopWords) || strlen($item) <= 3 ) {
2) Added a $count parameter to the function so that when it is called you can specify how many words to return.
1
2
3
4
5
6
7
function extractCommonWords($string,$count){
 
...
 
      $wordCountArr = array_slice($wordCountArr, 0, $count);
      return $wordCountArr;
}
The hardest thing to get right is the stop word list.

Brilliant. Thanks for this piece of code. I was after doing something very similar so have used this as a base and made a few tweaks here and there so it fits my needs. 

Hi

Great code and is working a treat on my site but I have a problem with words containing 'ss'.  The double ss is being removed so a word such as 'password' is being seen as 'paword' in the keywords.

Regards

 

Hi, would you please to let me know how I could include spanish characters too?.

For example I need to print: 'tamaño' instead I get: 'tamao'

Thanks in advance
Tolenca

 

philipnorton42's picture
Submitted by philipnorton42 on Wed, 06/22/2011 - 23:20

You need to add those special characters to the regular expression on line 6 of the examples above. You just need to make sure it takes alphanumeric characters as well as the spanish letters. I think that should work.

instead of

$wordCountArr = array();
1
2
3
4
5
6
7
8
9
10
11
12
if ( is_array($matchWords) ) {
          foreach ( $matchWords as $key =&gt; $val ) {
              $val = strtolower($val);
              if ( !isset($wordCountArr[$val]) {
                  $wordCountArr[$val] = array();
              }
              if ( isset($wordCountArr[$val]['count']) ) {
                  $wordCountArr[$val]['count']++;
              } else {
                  $wordCountArr[$val]['count'] = 1;
              }
          }

I did:

1
2
$ignoreOccur = array(1,2);
$wordCountArr = array_diff(array_count_values(explode(" ", matchWords)), $ignoreOccur);

To get all assoc array $wordCoundArr of [word] => [occurences], ignoring words that have occurred 1 or 2 (or specified number of) times.

Make sure this line:

1
$string = preg_replace('/ss+/i', '', $string);

reads like this:

1
$string = preg_replace('/\s\s+/i', '', $string);

This script just returns a string of comma separated keywords and supports multibyte characters.
Adjust your stopwords for your locale.

1
2
3
4
5
6
7
8
9
10
         function extractCommonWords($string)
        {       
                $stopwords = array('i','a','about','an','and','are','as','at','be','by','com','de','en','for','from','how','in','is','it','la','of','on','or','that','the','this','to','was','what','when','where','who','will','with','und','the','www');
                $string = mb_strtolower($string); // make it lowercase
                $string = trim(preg_replace('/\s\s+/i', '', $string;)); //remove multiple whitespace
                $string = preg_replace('/[\pP]/', '', $string); // remove punctuation
                $matchWords = array_filter(explode(&quot; &quot;,$string) , function ($item) use ($stopwords) { return !($item == '' || in_array($item, $stopwords) || mb_strlen($item) &lt; 2 || is_numeric($item));});
                $wordCountArr = arsort(array_count_values($matchWords));
                return implode(',', array_keys(array_slice($wordCountArr, 0, 10)));
        }

 

Whoops.. Ignore my previous code block.. it was bugged.. try this.

1
2
3
4
5
6
7
8
9
10
function extractKeyWords($string)
        {
                mb_internal_encoding('UTF-8');
                $stopwords = array('i','a','about','an','and','are','as','at','be','by','com','de','en','for','from','how','in','is','it','la','of','on','or','that','the','this','to','was','what','when','where','who','will','with','und','the','www');
                $string = preg_replace('/[\pP]/', '', trim(preg_replace('/\s\s+/i', '', mb_strtolower(utf8_encode($string)))));
                $matchWords = array_filter(explode(' ',$string) , function ($item) use ($stopwords) { return !($item == '' || in_array($item, $stopwords) || mb_strlen($item) &lt; 2 || is_numeric($item));});
                $wordCountArr = array_count_values($matchWords);
                arsort($wordCountArr);
                return implode(',', array_keys(array_slice($wordCountArr, 0, 10)));
        }

plz help me
how to use for arabic text such as "اب ت ث "?
thank you

philipnorton42's picture
Submitted by philipnorton42 on Thu, 05/17/2012 - 16:55

I'm afraid I can't answer that as I don't know enough about the Arabic language/alphabet. Sorry :(
Heya. Great script thanks :) Just a note for the guys who want to handle non-anglo characters, if they replace: $string = preg_replace('/[^a-zA-Z0-9 -]/', '', $string); with this: $string = preg_replace('/[^\w\d -]/', '', $string); It should support extended characters, not sure off the top of my head how much unicode is covered by \w (if any), but it works for a bunch of quick tests I've put through it.
to reduce whitespace (followed by another whitespace) within the string, its better to use preg_replace('/[ ]{2,}/sm', ' ', $text)
philipnorton42's picture
Submitted by philipnorton42 on Wed, 11/07/2012 - 23:32

@sammsel - Seems like a good way of doing things, but why is it better? Is there a performance overhead using my method? Also you are only detecting the space character, not all white space character, which might cause problems/

i see if $wordCountArr contain number (ex. year), the output value for it's 0.

For return the corect value you must modify

$wordCountArr = array_slice($wordCountArr, 0, 10);

in

$wordCountArr = array_slice($wordCountArr, 0, 10,$preserve_keys=true);

Thanks a lot for sharing this tutorial on how to extract keywords from a text string using php. This post really helped me big time. I am so glad that I came to this site. I just hope that these awesome posts will keep on coming.

Found this useful, I need a simple help. If there is a word "bmw-x666" It results "bmw,-,x666" but what i want is it shoudnt extract the word containing hyphen.

Could anyone help me?

How could i set the key phrase of matched words. eg. This is some, Vending Machines are

great code Philip - thanks !

is there a way to group the resulting keywords to two or three words ?

your eg: some,text,machines,vending

to eg: some text, machines vending, some machines, text vending, some vending, text machines... as multiple keywords can give good results in meta

working example: samsung, s3, battery, i9300, original, EB900F
result: samsung battery, samsung s3 battery, original i9300 battery, EB900F original, battery samsung s3...

google accepts up to 10 grouped keywords of one two or three keywords

Add new comment